Empirical scoring functions for structure-based virtual screening: applications, critical aspects, and challenges
Open accessAbstract
Structure-based virtual screening (VS) is a widely used approach that employs the knowledge of the three-dimensional structure of the target of interest in the design of new lead compounds from large-scale molecular docking experiments. Through the prediction of the binding mode and affinity of a small molecule within the binding site of the target of interest, it is possible to understand important properties related to the binding process. Empirical scoring functions are widely used for pose and affinity prediction. Although pose prediction is performed with satisfactory accuracy, the correct prediction of binding affinity is still a challenging task and crucial for the success of structure-based VS experiments. There are several efforts in distinct fronts to develop even more sophisticated and accurate models for filtering and ranking large libraries of compounds. This paper will cover some recent successful applications and methodological advances, including strategies to explore the ligand entropy and solvent effects, training with sophisticated machine-learning techniques, and the use of quantum mechanics. Particular emphasis will be given to the discussion of critical aspects and further directions for the development of more accurate empirical scoring functions.
AI snapshot
Written from the paper’s full text.
TL;DR
This review surveys empirical scoring functions for structure-based virtual screening, covering descriptors, training/test datasets, machine learning, and emerging strategies. It concludes that pose prediction is often satisfactory, but accurate binding-affinity prediction and ranking remain major challenges.
Key findings
- Pose prediction is generally satisfactory under adequate system preparation and when target flexibility is not significant, but active/inactive discrimination and absolute affinity prediction remain challenging; in D3R Grand Challenge 2, even with crystal structures of 36 complexes at Stage 2, the maximum Kendall’s tau achieved was 0.46.
- Linear scoring functions show moderate correlations (RP ∼ 0.6), whereas nonlinear machine-learning models achieve better correlations (RP > 0.7) on benchmarking studies; for example, SFCscoreRF reached R = 0.779, higher than linear SFCscore models.
- Key datasets include PDBbind-CN version 2017 with 17,900 structures (14,761 protein–ligand complexes) and affinities ranging from 1.2 pM to 10 mM; DUD-E has 102 targets with 50 decoys per ligand; DEKOIS 2.0 has 81 benchmarking sets for 80 targets with 30 decoys per active.
- Target-specific scoring functions can improve accuracy, but Wójcikowski et al. found RF target-specific models only slightly better than generic models, with two-thirds increasing EF1% by less than 10%.
Methods
- Design: narrative review of empirical scoring functions for structure-based virtual screening, including applications, critical aspects, and challenges.
- Data/sample: no primary experimental sample; synthesizes published literature, scoring functions, benchmarks (PDBbind-CN, DUD-E, DEKOIS 2.0, D3R Grand Challenges), and methodological studies.
- Analysis: qualitative synthesis of scoring-function components (descriptors, training/test sets, regression/classification, linear vs nonlinear machine learning) and evaluation metrics (docking, screening, ranking, and scoring power).
Limitations
- Paper states: there is no universal scoring function; affinity prediction remains challenging; nonlinear models act as “black boxes” and risk overfitting or bias; decoys are presumed rather than experimentally confirmed inactive; small benchmark performance differences (e.g., Spearman correlation differences of 0.05–0.15) lack statistical significance.
- Paper states remaining challenges in protein flexibility, solvation/water treatment, covalent docking, and QM/MM approaches; QM-based methods do not guarantee outperformance and still face solvent and entropic estimation problems.
- Evident from the text: as a narrative review, it does not conduct a systematic search or quantitative meta-analysis and reports no new primary data.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 150 cited sources checked · 0 supported · 0 flagged
- Premisecitation
Empirical scoring functions are developed to reproduce experimental affinity data based on the idea that it is possible to correlate the free energy of binding to a set of non-related variables.
“Empirical scoring functions are developed to reproduce experimental affinity data (Pason and Sotriffer, 2016) based on the idea that it is possible to correlate the free energy of binding to a set of non-related variables.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Pason and Sotriffer, 2016
Not checked in the free run.
- Rests on an assumption: Empirical scoring functions are widely used for pose and affinity prediction.
- Premisecitation
Empirical scoring functions have implemented specific terms accounting for intermolecular interactions, such as van der Waals and electrostatic potentials.
“Empirical scoring functions have implemented specific terms accounting for intermolecular interactions, such as van der Waals and electrostatic potentials.”
Quote in the paperNot a substantial claim- Rests on an assumption: Empirical scoring functions are widely used for pose and affinity prediction.
- Premisecitation
Current docking methods and associated scoring functions exhibit good pose prediction power if the system is adequately prepared and target flexibility does not play a significant role.
“Current docking methods and the associated scoring functions exhibit good pose prediction power if one assumes an adequate preparation of the system and if the target flexibility does not play a significant role (Corbeil et al., 2012; Chaput and Mouawad, 2017).”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites Corbeil et al., 2012 source
Not checked in the free run.
- cites Chaput and Mouawad, 2017
Not checked in the free run.
- Interpretationreasoning
The inclusion of a polar desolvation might be crucial to avoid overestimation of hydrogen bonds, since H-bond formation is directly related with the desolvation of polar atoms.
“Besides the improvement in affinity predictions, the inclusion of a polar desolvation might be crucial to avoid overestimation of hydrogen bonds, since the H-bond formation is directly related with the desolvation of polar atoms.”
Quote in the paperNot a substantial claim - Interpretationreasoning
Metal ions can be a source of inaccuracy when using non-specific scoring functions, since the real contribution of metal-ion interactions can be underestimated or overestimated.
“Despite the importance in considering metal ions, it can be also a source of inaccuracy when using non-specific scoring functions, since the real contribution of interaction metal ions can be underestimated – in the case of simple counting of metal-atom interacting pairs – or overestimated – when using Coulomb potential with formal charges.”
Quote in the paperNot a substantial claim
- Premisereasoning
Empirical models differ in the number and type of descriptors, the algorithm adopted for training, and the number, diversity, and quality of protein-ligand complexes used during parameterization.
“The empirical models differ in the number and type of descriptors; the algorithm adopted for training the model; and the number, the diversity, and the quality data of protein–ligand complexes used during the parameterization process.”
Quote in the paperNot a substantial claim - Premisecitation
The desolvation of polar or charged groups of the protein or ligand is unfavorable to binding when the formed solute-solvent interactions are not effectively satisfied upon protein-ligand binding.
“At the same time, the desolvation of polar or charged groups of the protein or ligand is unfavorable to binding when the formed solute–solvent interactions are not effectively satisfied upon the protein–ligand binding (Blaber et al., 1993; Kar et al., 2013).”
Quote in the paperNot a substantial claim0 of 2 sources checked - Premisecitation
The conjunction of nonlinear models and more straightforward atom counting descriptors is subject to significant criticisms.
“However, the conjunction of nonlinear models and more straightforward atom counting descriptors is subjected to significant criticisms (Gabel et al., 2014).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Gabel et al., 2014 source
Not checked in the free run.
- Rests on an assumption: Empirical scoring functions are widely used for pose and affinity prediction.
- Premisenone given
The restriction of the rotatable bonds of the ligand after formation of the protein-ligand complex promotes an entropic loss that is unfavorable to binding affinity.
“The restriction of the rotatable bonds of the ligand after the formation of the protein–ligand complex also promotes an entropic loss (Sconf) that is unfavorable to the binding affinity.”
Quote in the paperNot a substantial claim - Premisecitation
The energy gap and the number of local binding wells extracted from multiple binding modes can distinguish true binders among high-scored decoys.
“For example, Wei et al. (2010) developed two new parameters extracted from the multiple binding modes, generated by the AutoDock 3.05 program, and combined them for classification purposes using logistic regression to distinguish true binders among high-scored decoys.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Wei et al., 2010 source
Not checked in the free run.
- Premisereasoning
Many scoring functions have implemented desolvation terms to introduce the hydrophobic effect and/or penalize buried and not interacting polar/charged atoms after protein-ligand binding to improve binding affinity predictions.
“In this context, many scoring functions have implemented desolvation terms to introduce the hydrophobic effect and/or penalize buried and not interacting polar/charged atoms after protein–ligand binding to improve binding affinity predictions.”
Quote in the paperNot a substantial claim - Premisereasoning
Decoys in DUD-E are presumed, not experimentally verified, to be inactive compounds.
“Decoys are presumed, not experimentally verified, to be inactive compounds since they are chosen to be topologically distinct from ligands but exhibiting similar physicochemical properties.”
Quote in the paperSubstantial claim - Interpretationreasoning
The use of benchmarking sets such as DUD-E and DEKOIS2.0 is directly dependent on the docking program adopted because experimental structures of the protein-ligand complexes are not available as in PDBbind datasets.
“The use of benchmarking sets such as DUD-E and DEKOIS2.0 is directly dependent on the docking program adopted since the experimental structures of the protein–ligand complexes are not available as in the PDBbind datasets.”
Quote in the paperNot a substantial claim - Premisecitation
DEKOIS 2.0 is composed of 81 benchmarking sets for 80 protein targets of therapeutic relevance, including nonconventional targets such as protein-protein interaction complexes.
“DEKOIS 2.0 is composed of 81 benchmarking sets for 80 protein targets of therapeutic relevance, including nonconventional targets such as protein–protein interaction complexes (Bauer et al., 2013).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Bauer et al., 2013 source
Not checked in the free run.
- Interpretationcitation
The use of decoys instead of validated inactive compounds remains a major drawback for most datasets since no experimental activity is reported for them and confirmed inactive molecules are scarce.
“The use of decoys instead of validated inactive compounds remains a major drawback for most datasets since no experimental activity are reported for them, and the number of confirmed inactive molecules is too scarce (Lagarde et al., 2015; Chaput et al., 2016b; Réau et al., 2018).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites Lagarde et al., 2015 source
Not checked in the free run.
- cites Chaput et al., 2016b
Not checked in the free run.
- cites Réau et al., 2018
Not checked in the free run.
- Interpretationreasoning
Scoring function training or validation in VS experiments using DUD-E and DEKOIS2.0 is performed with no warranty that the ligand poses were correctly predicted.
“In fact, the scoring function training or validation in VS experiments using these datasets is performed with no warranty that the ligand poses were correctly predicted.”
Quote in the paperSubstantial claim - Premisecitation
There is a tradeoff between the size of the training and validation/test sets.
“There is a tradeoff between the size of the training and validation/test sets.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Abu-Mostafa et al., 2012
Not checked in the free run.
- Premisecitation
The initial automatic preparation of structures performed by PDBbind did not provide an optimized hydrogen bond network and appropriate assignment of protonation/tautomeric states of the α-amylase and MeG2-GHIL complex.
“For example, the initial automatic preparation of the structures performed by PDBbind did not provide an optimized hydrogen bond network and appropriate assignment of protonation/tautomeric states of the α-amylase and MeG2-GHIL complex [Figure 1, PDB code 1U33; Numao et al., 2004].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Numao et al., 2004 source
Not checked in the free run.
- Premisecitation
Even with crystal structures of 36 complexes at Stage 2 of Grand Challenge 2, the maximum Kendall's tau achieved was 0.46.
“Even with the crystal structures of 36 complexes at Stage 2, the maximum Kendall’s tau achieved was 0.46, reinforcing the great deal in correctly ranking a set of compounds.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Gaieb et al., 2018 source
Not checked in the free run.
- Interpretationreasoning
Preparation of protein-ligand complexes has a direct influence on training and evaluation of scoring functions, mainly for scoring functions based on force-field descriptors.
“Indeed, the preparation of protein–ligand complexes has a direct influence on training and evaluation of scoring functions, mainly for scoring functions based on force-field descriptors.”
Quote in the paperSubstantial claim - Premisecitation
Since most benchmarking studies evaluate scoring functions on a few hundred complexes, small differences in Spearman correlation coefficient between 0.05 and 0.15 lack statistical significance.
“Since most benchmarking studies evaluate scoring functions on a few hundred complexes, small differences in Spearman correlation coefficient between 0.05 and 0.15, for example, lack statistical significance (Carlson, 2013, 2016).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Carlson, 2013, 2016
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Premisecitation
The predictive performance of scoring functions may vary between different benchmarking experiments due to factors such as composition of the dataset, structural quality of the complexes, level of experience of the researchers, and protocol of preparation of the complexes.
“The predictive performance of scoring functions may vary between different benchmarking experiments due to factors such as: (i) composition of the dataset, (ii) structural quality of the complexes, (iii) level of experience of the researches performing the experiments, and (iv) protocol of preparation of the complexes (Yuriev and Ramsland, 2013).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Yuriev and Ramsland, 2013
Not checked in the free run.
- Premisecitation
Correct assignment of protein and ligand protonation/tautomeric states with respect to experimental pH, Asn/Gln/His flips, and defined stereocenters is crucial and requires careful inspection of structures.
“The correct assignment of both protein and ligand protonation/tautomeric states with respect to the experimental pH, Asn/Gln/His flips, and defined stereocenters of the compounds are crucial, requiring a careful inspection of the structures (Kalliokoski et al., 2009; Martin, 2009; Petukh et al., 2013; Sastry et al., 2013).”
Quote in the paperSubstantial claim0 of 3 sources checked - Interpretationcitation
Small differences in calculated performances are generally insufficient to state which scoring function performs better when comparing top-ranked models.
“Although ranking scoring functions according to their performances for affinity prediction on benchmark sets highlights the more competitive models, it is important to observe that small differences in the calculated performances are generally insufficient to state which scoring function performs better than other when comparing the top-ranked models.”
Quote in the paperSubstantial claim - Premisecitation
Little attention has been paid to careful preparation of protein-ligand structures, usually relying on automatic procedures.
“Despite many efforts made for collecting even more extensive and better quality datasets, little attention has been paid to the careful preparation of the protein–ligand structures, usually relying on automatic procedures (Bauer et al., 2013).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Bauer et al., 2013 source
Not checked in the free run.
- Interpretationreasoning
Larger benchmarking sets composed of high-quality protein-ligand complex structures are required for a reliable comparison of docking methodologies and scoring functions.
“Thus, larger benchmarking sets composed of high-quality protein–ligand complexes structures are required for a reliable comparison of docking methodologies and scoring functions.”
Quote in the paperNot a substantial claim - Premisecitation
It is possible to use both classification and regression approaches to deal with the same problem of binding affinity prediction.
“It is also possible to use both classification and regression approaches to deal with the same problem of binding affinity prediction.”
Quote in the paperNot a substantial claim - Interpretationcitation
The superiority of nonlinear models has been confirmed through the rebuild of linear scoring functions using nonlinear algorithms.
“The superiority of nonlinear models has also been confirmed through the rebuild of linear scoring functions using nonlinear algorithms, i.e., scoring functions trained with the same original descriptors of the correspondent linear model but with a different regression method.”
Quote in the paperSubstantial claim - Premisereasoning
The most crucial difference between linear and nonlinear scoring functions is that linear requires a predefined functional form, whereas nonlinear implicitly derives the mathematical relationship between descriptors, allowing combination of variables and higher order exponents.
“The most crucial difference between linear and nonlinear scoring functions is that the former requires a predefined functional form (e.g., the sum of terms in the case of linear scoring functions), whereas the latter implicitly derives the mathematical relationship between the descriptors, allowing the combination of variables and higher order exponents for the terms.”
Quote in the paperNot a substantial claim - Premisecitation
The use of decoys instead of confirmed inactive compounds has important impacts in training and measuring the performance of scoring functions.
“Furthermore, the use of decoys instead of confirmed inactive compounds has important impacts in training and measuring the performance of scoring functions (Chaput et al., 2016b; Réau et al., 2018).”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites Chaput et al., 2016b
Not checked in the free run.
- cites Réau et al., 2018
Not checked in the free run.
- Interpretationreasoning
For affinity prediction purposes, the use of datasets with curated affinity data is essential for reliable predictions and benchmarking.
“Especially for affinity prediction purposes, the use of datasets with curated affinity data is essential for reliable predictions and benchmarking.”
Quote in the paperSubstantial claim - Premisecitation
It is possible to use scoring functions trained with regression methods to classify active and inactive molecules given a predetermined range of affinity data for defining active and inactive compounds.
“It is possible to use scoring functions trained with regression methods to classify active and inactive molecules given a predetermined range of affinity data for defining active and inactive compounds (Ain et al., 2015).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Ain et al., 2015 source
Not checked in the free run.
- Premisecitation
Many factors such as inherent experimental error can be a source of inaccuracies, limiting the average prediction error achievable on large datasets.
“Despite the efforts in collecting high-quality affinity data, many factors such as the inherent experimental error can be a source of inaccuracies, limiting the average prediction error achievable on large datasets (Shoichet, 2006; Ferreira et al., 2009; Sotriffer and Matter, 2011; Kramer et al., 2012).”
Quote in the paperNot a substantial claim0 of 3 sources checked - Premisecitation
Li et al. investigated the replacement of MLR by RF for regression using the same Cyscore descriptors and found that the nonlinear model improved affinity prediction.
“Li et al. (2014a) investigated the replacement of MLR by RF for regression using the same Cyscore descriptors and found that the nonlinear model improved the affinity prediction.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Li et al. (2014a)
Not checked in the free run.
- Premisecitation
Pason and Sotriffer used a strategy of classifying complexes using algorithms such as KNN and further generating linear regression models for each cluster, achieving predictive performances comparable to a nonlinear scoring function trained with RF.
“For example, Pason and Sotriffer (2016) used a strategy of classifying the complexes using algorithms such as KNN and further generating linear regression models for each cluster achieving predictive performances comparable to that obtained by the nonlinear scoring function trained with RF.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Pason and Sotriffer (2016)
Not checked in the free run.
- Interpretationcitation
The advantage of nonlinear scoring functions partially circumvents the problematic modeling assumptions of linear models.
“This advantage of nonlinear scoring functions partially circumvents the problematic modeling assumptions of linear models (Dill, 1997; Baum et al., 2010; Sotriffer, 2012).”
Quote in the paperNot a substantial claim0 of 3 sources checked - Interpretationcitation
The development of scoring functions adapted for flexible receptor docking is crucial to achieve real improvements in pose and affinity prediction.
“In this context, the development of scoring functions adapted for flexible receptor docking is crucial to achieve real improvements in pose and affinity prediction (Totrov and Abagyan, 1997; Wei et al., 2002; Fischer et al., 2014; Ravindranath et al., 2015; Lam et al., 2017; Kong et al., 2018).”
Quote in the paperSubstantial claim0 of 3 sources checked - Interpretationcitation
The main drawback of nonlinear scoring functions is that they work as "black boxes" since the relationship between descriptors is often vague, requiring careful use to avoid meaningless interpretations.
“Despite nonlinear scoring functions have the main advantage of discarding the necessity of a pre-defined functional form, their main drawback is that they work as “black boxes” since the relationship between the descriptors is often vague, requiring careful use to avoid meaningless interpretations (Gabel et al., 2014).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Gabel et al., 2014 source
Not checked in the free run.
- Interpretationcitation
Nonlinear models, together with the use of a significant amount of descriptors lacking physical meaning, offer the risk of producing excellent performance indexes due to overfitting and/or bias to the training set construction.
“Together with the use of a significant amount of descriptors lacking physical meaning, nonlinear models offer the risk of producing excellent performance indexes due to overfitting and/or bias to the training set construction (e.g., capturing the rules adopted during the selection of active and decoy compounds) (Hawkins, 2004; Abu-Mostafa et al., 2012).”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites Hawkins, 2004 source
Not checked in the free run.
- cites Abu-Mostafa et al., 2012
Not checked in the free run.
- Premisecitation
Fischer et al. successfully identified new ligands targeting specific receptor conformations of cytochrome c peroxidase using a flexible docking method that samples and weights protein conformations guided by experimentally derived conformations, integrating Boltzmann-weighted energy penalties related with protein flexibility to the DOCK3.7 scoring function.
“Recently, Fischer et al. (2014) successfully identified new ligands targeting specific receptor conformations of cytochrome c peroxidase using a flexible docking method that samples and weights protein conformations guided by experimentally derived conformations, integrating the Boltzmann-weighted energy penalties related with the protein flexibility to the DOCK3.7 scoring function.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Fischer et al. (2014) source
Not checked in the free run.
- Interpretationcitation
Protein flexibility is still a great challenge for docking programs and scoring functions.
“Protein flexibility is still a great challenge for docking programs and scoring functions (Cavasotto and Singh, 2008; Tuffery and Derreumaux, 2012; Buonfiglio et al., 2015; Spyrakis and Cavasotto, 2015; Kurkcuoglu et al., 2018).”
Quote in the paperSubstantial claim0 of 3 sources checked- cites Cavasotto and Singh, 2008
Not checked in the free run.
- cites Tuffery and Derreumaux, 2012
Not checked in the free run.
- cites Buonfiglio et al., 2015 source
Not checked in the free run.
- Premisereasoning
Most docking methodologies adopt a single, rigid conformation of the receptor due to high computational cost and methodological limitations proportional to the increase in flexibility.
“Most docking methodologies adopt a single, rigid conformation of the receptor, due to the high computational cost and methodological limitations proportional to the increase in the degree of flexibility.”
Quote in the paperNot a substantial claim - Premisecitation
Pason and Sotriffer demonstrated that it is possible to achieve similar high performances of nonlinear models through the development of a set of linear scoring functions trained using clustered - smaller and more homogeneous - datasets of protein-ligand complexes.
“Pason and Sotriffer (2016) demonstrated that it is possible to achieve similar high performances of nonlinear models through the development of a set of linear scoring functions trained using clustered – smaller and more homogeneous – datasets of protein–ligand complexes.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Pason and Sotriffer (2016)
Not checked in the free run.
- Premisecitation
Water molecules play an essential role in the ligand-protein binding process.
“Water molecules play an essential role in the ligand–protein binding process.”
Quote in the paperSubstantial claim - Premisecitation
Some docking engines have implemented strategies to treat water molecules explicitly with adapted scoring functions once water molecules are identified.
“Once the water molecules are identified, some docking engines have implemented strategies to treat water molecules explicitly with adapted scoring functions.”
Quote in the paperNot a substantial claim - Interpretationcitation
Despite many efforts to include protein flexibility in VS experiments, the complex and multifactorial framework of flexible protein-ligand binding is still a great challenge.
“Despite the many efforts made to include the protein flexibility in VS experiments, the complex and multifactorial framework of flexible protein–ligand binding is still a great challenge (Bottegoni et al., 2011; Nunes-Alves and Arantes, 2014; Antunes et al., 2015; Buonfiglio et al., 2015; Kong et al., 2018).”
Quote in the paperSubstantial claim0 of 3 sources checked - Interpretationcitation
The correct prediction of the free energy of binding associated with the ligand displacement of water molecules is a key challenge for currently available docking scoring functions.
“The correct prediction of the free energy of binding associated to the ligand displacement of water molecules is a key challenge for the currently available docking scoring functions (Riniker et al., 2012; Spyrakis and Cavasotto, 2015; Bodnarchuk, 2016).”
Quote in the paperSubstantial claim0 of 3 sources checked - Premisecitation
Individual water molecules can stabilize the ligand binding mode through the formation of water bridges or a water-mediated hydrogen-bond network.
“Besides the hydrophobic and desolvation effects, individual water molecules can stabilize the ligand binding mode through the formation of water bridges or a water-mediated hydrogen-bond network (Poornima and Dean, 1995; Levy and Onuchic, 2006).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Poornima and Dean, 1995
Not checked in the free run.
- cites Levy and Onuchic, 2006
Not checked in the free run.
- Premisecitation
In WScore, the location and thermodynamics of explicit waters are predicted using WaterMap and integrated to the scoring function together with a desolvation term to penalize associated desolvation of polar or uncharged groups of protein or ligand.
“For instance, in the WScore docking methodology, the location and thermodynamics of explicit waters are predicted using WaterMap and integrated to the scoring function together with a desolvation term to penalize the associated desolvation of polar or uncharged groups of protein or ligand (Murphy et al., 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Murphy et al., 2016 source
Not checked in the free run.
- Premisecitation
There is a whole class of inhibitors that form a covalent bond with their enzyme/receptor target.
“However, there is a whole class of inhibitors that form a covalent bond with their enzyme/receptor target (De Cesco et al., 2017).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites De Cesco et al., 2017 source
Not checked in the free run.
- Premisecitation
The GOLD program considers an all-atom and flexible water model able to rotate around its three principal axes, and rewards water displacement in the GoldScore or ChemScore scoring functions according to a balance between loss of rigid-body entropy and change in interaction energies on binding to the protein cavity.
“The GOLD program considers all-atom and flexible water model able to rotate around its three principal axes, and rewards water displacement in the GoldScore or ChemScore scoring functions according to a balance between the loss of rigid-body entropy and the change in the interaction energies on binding to the protein cavity (Verdonk et al., 2005).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Verdonk et al., 2005 source
Not checked in the free run.
- Interpretationcitation
QM methods are particularly useful when molecular recognition involves bond formation, π-stacking, cation-π, halogen bonding, and polarization and charge transfer effects.
“This is particularly true when dealing with systems where the molecular recognition involves bond formation, π-stacking, cation-π, halogen bonding (i.e., σ-hole bonding), and polarization and charge transfer effects (Christensen et al., 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Christensen et al., 2016 source
Not checked in the free run.
- Premisecitation
GOLD, Autodock4, CovalentDock, CovDock, DOCKovalent, and DOCK-TITE are examples of docking programs that developed specific methodologies to deal with covalent docking.
“GOLD (Jones et al., 1997), Autodock4 (Bianco et al., 2016), CovalentDock (Ouyang et al., 2013), CovDock (Zhu et al., 2014), DOCKovalent (London et al., 2014), and DOCK-TITE (Scholz et al., 2015) are some examples of docking programs that developed specific methodologies to deal with covalent-docking.”
Quote in the paperNot a substantial claim0 of 3 sources checked - Premisecitation
Results obtained using QM or hybrid QM/MM-based methods are very encouraging compared to standard scoring functions, principally when dealing with metalloproteins.
“The results obtained using QM or hybrid QM/MM-based methods are very encouraging when compared to the standard scoring functions, principally when dealing with metalloproteins (Chaskar et al., 2017; Pecina et al., 2018).”
Quote in the paperSubstantial claim0 of 2 sources checked - Premisecitation
Consensus scoring works due to a simple statistical reason related to the law of large numbers: the mean value found by repeated independent predictions tends toward the real and expected value.
“They demonstrated that CS works due to a simple statistical reason related to the law of large numbers: the mean value found by repeated independent predictions tends toward the real and expected value.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Wang and Wang, 2001
Not checked in the free run.
- Interpretationcitation
Results from the 2016 D3R Grand Challenge indicate that the use of QM/MM scoring could be a powerful strategy.
“Moreover, the results from the 2016 D3R Grand Challenge indicate that the use of QM/MM scoring could be a powerful strategy (Gao et al., 2018).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Gao et al., 2018 source
Not checked in the free run.
- Interpretationcitation
Target-specific scoring functions could be more efficient in accounting for specific interactions and particular binding characteristics associated with a target class of interest.
“Hence, it is expected that they could be more efficient in accounting for specific interactions and particular binding characteristics associated with a target class of interest (Seifert, 2009).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Seifert, 2009 source
Not checked in the free run.
- Rests on an assumption: Target-specific scoring functions yield significant improvements in docking and VS accuracies compared to non-specific models when trained on target-specific protein–ligand complexes.
- Premisecitation
Combining different scoring functions in a consensus scoring scheme is a promising data fusion strategy to improve VS enrichment, pose, and affinity prediction.
“The combination of different scoring functions on a scoring scheme (CS) is considered as a promising data fusion strategy to improve VS enrichment, pose, and affinity prediction (Charifson et al., 1999; Bissantz et al., 2000; Yang et al., 2005; Kaserer et al., 2015; Chaput et al., 2016a; Chaput and Mouawad, 2017; Ericksen et al., 2017).”
Quote in the paperSubstantial claim0 of 3 sources checked - Premisecitation
Wang and Wang (2001) provided a theoretical basis for the effectiveness of consensus scoring on affinity prediction.
“In the very interesting work of Wang and Wang (2001), they provided a theoretical basis for the effectiveness of CS on affinity prediction.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wang and Wang, 2001
Not checked in the free run.
- Premisecitation
Rebuilding the AutoDock4 scoring function using ligand partial charges calculated with QM methods and protein charges from Amber99SB instead of the Gasteiger method improved both pose and affinity predictions.
“Wang et al. (2011) rebuild the AutoDock4 scoring function using ligand partial charges calculated with QM methods and protein charges from the Amber99SB instead of the Gasteiger method, improving both pose and affinity predictions.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wang et al., 2011 source
Not checked in the free run.
- Interpretationcitation
The consensus scoring strategy could overcome to some extent the limitations faced by the single-scoring approach, for example, inconsistent performances across different protein targets and chemical classes.
“The CS strategy could overcome to some extent the limitations faced by the single-scoring approach, for example, the inconsistent performances across different protein targets and chemical classes (Moitessier et al., 2009).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Moitessier et al., 2009 source
Not checked in the free run.
- Interpretationcitation
QM-based approaches still face the same problems associated with the correct estimation of solvent and other entropic effects to the protein-ligand binding free energy.
“and they still face the same problems associated with the correct estimation of the solvent and other entropic effects to the protein–ligand binding free energy.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Crespo et al., 2017 source
Not checked in the free run.
- Premisecitation
Nonlinear consensus scoring models were developed to improve pose prediction and ranking of compounds in VS experiments.
“Nonlinear CS models were also developed to improve pose prediction and ranking compounds in VS experiments (Betzi et al., 2006; Teramoto and Fukunishi, 2007; Ashtawy and Mahapatra, 2015; Ericksen et al., 2017).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites Betzi et al., 2006 source
Not checked in the free run.
- cites Teramoto and Fukunishi, 2007
Not checked in the free run.
- cites Ashtawy and Mahapatra, 2015
Not checked in the free run.
- Premisecitation
In prospective VS studies against cyclooxygenases 1 and 2, the chance of a compound being truly active increases when more tools predicted it as active.
“For instance, Kaserer et al. (2015) applied CS on prospective VS studies against cyclooxygenases 1 and 2 and found that the chance of a compound to be truly active increases when more tools predicted it as active.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Kaserer et al., 2015 source
Not checked in the free run.
- Interpretationcitation
It is not guaranteed that QM-based approaches will always outperform standard scoring functions.
“Nevertheless, it is important to note that it is not guaranteed that QM-based approaches will always outperform standard scoring functions (Crespo et al., 2017)”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Crespo et al., 2017 source
Not checked in the free run.
- Premisecitation
A scoring function specific to Heat Shock Protein 90 (HSP90) was successfully designed and applied in VS.
“In a more recent work, a scoring function specific to Heat Shock Protein 90 (HSP90) was successfully designed and applied in VS (Santos-Martins, 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Santos-Martins, 2016 source
Not checked in the free run.
- Premisecitation
Adapting the Fresno empirical scoring function to the class I MHC HLA-B*2705 protein significantly improved affinity prediction over six different traditional scoring functions.
“For instance, Logean et al. (2001) adapted the Fresno empirical scoring function to the class I MHC HLA-B∗2705 protein with a significant improvement in affinity prediction over six different traditional scoring functions.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Logean et al., 2001 source
Not checked in the free run.
- Premisecitation
HADDOCKPPI is a linear scoring function specifically developed to predict binding affinities of inhibitors of protein-protein interactions (iPPIs).
“The HADDOCKPPI is a linear scoring function specifically developed to predict binding affinities of inhibitors of protein–protein interactions (iPPIs)”
Quote in the paperNot a substantial claim0 of 2 sources checked - Premisecitation
Nonlinear scoring functions specific for protein classes/targets achieved superior performance than generic models.
“In general, nonlinear scoring functions specific for protein classes/targets also achieved superior performance than the generic models (Wang et al., 2015; Ashtawy and Mahapatra, 2018).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Wang et al., 2015 source
Not checked in the free run.
- cites Ashtawy and Mahapatra, 2018
Not checked in the free run.
- May combine two claims. 42
- Classified as probably not a substantial claim. 37
May combine two claims.
PremisecitationThe nonpolar desolvation is favorable to binding and is related to the hydrophobic effect when transferring nonpolar molecular surface from bulk water to a nonpolar medium.
“The nonpolar desolvation, favorable to binding, is related to the hydrophobic effect when transferring nonpolar molecular surface from the bulk water to a medium that is nonpolar, as is the case of many protein binding cavities (Tanford, 1980; Williams and Bardsley, 1999; Freire, 2008).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites Tanford, 1980
Not checked in the free run.
- cites Williams and Bardsley, 1999
Not checked in the free run.
- cites Freire, 2008 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationScoring functions are typically divided into three main classes: force field-based, knowledge-based, and empirical.
“Scoring functions are typically divided into three main classes (Wang et al., 2003): force field-based, knowledge-based, and empirical.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wang et al., 2003 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMany studies have highlighted the influence of halogen bonds on enhancing binding affinity against several targets.
“Many studies have highlighted the influence of halogen bonds (X-bonds) on enhancing binding affinity against several targets and the computational methods developed so far (Desiraju et al., 2013; Ford and Ho, 2016).”
Quote in the paperNot a substantial claim0 of 2 sources checkedMay combine two claims.
PremisecitationScoring functions are a straightforward and fast strategy despite limited accuracy, remaining the main alternative to be applied in VS experiments.
“In this context, scoring functions emerge as a straightforward and fast strategy despite limited accuracy, remaining as the main alternative to be applied in VS experiments (Huang et al., 2010).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Huang et al., 2010 source
Not checked in the free run.
- Rests on an assumption: Empirical scoring functions are widely used for pose and affinity prediction.
May combine two claims.
PremisecitationDetection of active compounds among decoys and accurate prediction of binding affinity remain challenging tasks, even when induced fit and entropy effects are not important for binding.
“However, the detection of active compounds among a set of decoy compounds and the accurate prediction of binding affinity remain challenging tasks, even when induced fit and entropy effects are not important for binding (Gohlke and Klebe, 2002; Damm-Ganamet et al., 2013; Yuriev and Ramsland, 2013; Grinter and Zou, 2014; Smith et al., 2016).”
Quote in the paperSubstantial claim0 of 3 sources checked- cites Gohlke and Klebe, 2002
Not checked in the free run.
- cites Damm-Ganamet et al., 2013 source
Not checked in the free run.
- cites Yuriev and Ramsland, 2013
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationThe development of more accurate scoring functions is strategic in the field of SBDD and remains a challenging task, especially in hit-to-lead optimization and de novo design.
“Moreover, the development of more accurate scoring functions is strategic in the field of SBDD and remains a challenging task, especially in the hit-to-lead optimization (Enyedy and Egan, 2008) and de novo design (Liu et al., 2017).”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites Enyedy and Egan, 2008
Not checked in the free run.
- cites Liu et al., 2017 source
Not checked in the free run.
- Rests on an assumption: There is no universal scoring function with significant reliability for all molecular systems.
Probably not a substantial claim: may not pass S3 (load-bearing).
Premisenone givenDUD-E keeps a proportion of 50 decoys per ligand for all 102 macromolecules.
“Although the number of ligands (i.e., active compounds) significantly varies for each target, a proportion of 50 decoys per ligand is kept for all 102 macromolecules.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationMM-GBSA is less sensitive to the investigated systems and is more suitable to be used in general cases than MM-PBSA.
“They concluded that although similar results were observed, MM-GBSA is less sensitive to the investigated systems and is more suitable to be used in general cases (e.g., reverse docking, which is widely used to predict the receptor target(s) of a compound).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Sun et al., 2014 source
Not checked in the free run.
May combine two claims.
PremisecitationDUD-E is an enhanced version of the original DUD set and has been widely used to train and validate scoring functions.
“The DUD-E dataset is an enhanced version of the original DUD set and has been widely used to train and validate scoring functions (Huang et al., 2006; Mysinger et al., 2012).”
Quote in the paperNot a substantial claim0 of 2 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Interpretationnone givenA significant drawback of PDBbind-CN datasets is the insufficient information regarding negative data.
“A significant drawback of PDBbind-CN datasets is the insufficient information regarding negative data (i.e., experimentally confirmed inactive compounds).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Premiseown dataPDBbind-CN collected affinities range from 1.2 pM to 10 mM.
“Collected affinities comprise a large interval of values, ranging from 1.2 pM (1.2 × 10−12 M) to 10 mM (1.0 × 10−3 M).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationDUD-E is composed of 102 targets with corresponding active, inactive, marginal, and decoy compounds.
“It is composed of 102 targets with corresponding active, inactive, marginal, and decoy compounds.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Mysinger et al., 2012 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisereasoningThe PDBbind-CN refined set is considered one of the largest datasets of structures available for the development and validation of docking methodologies and scoring functions.
“The “refined set” is a subset composed of high-quality datasets constructed according to several criteria concerning the quality of the structures, the affinity data, and the nature of the complex, being considered one of the largest datasets of structures available for the development and validation of docking methodologies and scoring functions.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationThe unfavorable desolvation effect from burying polar groups after ligand binding plays an important role in the binding event, but it is commonly neglected by most scoring functions.
“The unfavorable desolvation effect from burying polar groups after ligand binding also plays an important role in the binding event, but it is commonly neglected by most scoring functions (Kar et al., 2013; Li et al., 2014c; Cramer et al., 2017).”
Quote in the paperNot a substantial claim0 of 3 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe PDBbind-CN core set is widely used for comparative assessment of scoring functions in predicting affinities.
“Also, PDBbind-CN provides a benchmarking named “core set” widely used for comparative assessment of scoring functions in predicting affinities (Li Y. et al., 2018).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Li Y. et al., 2018
Not checked in the free run.
May combine two claims.
PremisecitationScoring functions generally neglect entropic contributions or adopt simplified algorithms to approximate entropies.
“Given the difficulty in modeling entropic effects for ΔGbind, scoring functions generally neglect their contributions or adopt simplified algorithms to approximate entropies in a straightforward manner (Jain, 2006).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Jain, 2006 source
Not checked in the free run.
- Rests on an assumption: Empirical scoring functions are widely used for pose and affinity prediction.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationPDBbind-CN current release (version 2017) contains 17,900 structures, including 14,761 protein-ligand complexes.
“The current release (version 2017) contains 17,900 structures (14,761 protein–ligand complexes) and is annually updated to keep up with the growth of the Protein Data Bank (Berman et al., 2000).”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites Liu et al., 2015
Not checked in the free run.
- cites Berman et al., 2000 source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based).
Study design and conductnone givenThe test set must be composed of instances not used in the training process at any moment.
“Furthermore, the test set must be composed of instances not used in the training process at any moment.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationIn Wójcikowski et al. (2017), fivefold cross-validations using the DUD-E dataset considered horizontal, vertical, and per-target splitting strategies.
“For example, in the recent work of Wójcikowski et al. (2017), they performed fivefold cross-validations using the DUD-E dataset. Three distinct splitting strategies were considered: horizontal, vertical, and per-target.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wójcikowski et al., 2017 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationIn the per-target split, training and test are performed for each protein target, simulating the construction and validation of target-specific scoring functions.
“in the per-target split, the training and test are performed for each protein target (i.e., 102 unique machine-learning models relative to the 102 DUD-E targets), simulating the construction and validation of target-specific scoring functions.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wójcikowski et al., 2017 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationPDBbind, DUD-E, and DEKOIS 2.0 are examples of widely used benchmarks for evaluating scoring functions.
“PDBbind (Liu et al., 2015), DUD-E (Mysinger et al., 2012), and DEKOIS 2.0 (Bauer et al., 2013) are examples of widely used benchmarks for evaluating scoring functions.”
Quote in the paperNot a substantial claim0 of 3 sources checkedMay combine two claims.
PremisecitationStudies evaluating the influence of training size demonstrated that MLR becomes insensitive to the growth of the training size whereas larger training sets can lead to an overall better accuracy of nonlinear scoring functions.
“Studies evaluating the influence of the training size for the performance of linear and nonlinear scoring functions for affinity prediction demonstrated that MLR becomes insensitive to the growth of the training size whereas larger training sets can lead to an overall better accuracy of nonlinear scoring functions (Ding et al., 2013; Ain et al., 2015; Li et al., 2015a,b; Li H. et al., 2018).”
Quote in the paperNot a substantial claim0 of 3 sources checkedProbably not a substantial claim: may not pass S4 (based).
Study design and conductnone givenTraining, validation, and test sets must never have protein-ligand complexes in common at the same time.
“It is important to keep in mind that training, validation, and test sets must never have protein–ligand complexes in common at the same time.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim. May combine two claims.
Study design and conductreasoningThe test set must be used only for evaluating the predictive performance of different scoring functions, and no decision should be taken based on its performance to avoid artificially high correlations.
“Thus, the test set must be used only for evaluating the predictive performance of different scoring functions, and no decision should be taken based on the performance for this dataset to avoid useless comparisons due to artificially high correlations.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationHuang and Wong developed the screening performance index (SPI) to evaluate VS methods that correlate with BEDROC with less computational cost, discarding the necessity of docking decoy compounds.
“Recently, Huang and Wong (2016) developed an inexpensive method – the screening performance index (SPI) – to evaluate VS methods that correlate with BEDROC with less computational cost, since it discards the necessity of docking decoy compounds (i.e., only considers the docking of active molecules).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Huang and Wong (2016)
Not checked in the free run.
May combine two claims.
PremisecitationAccording to results from Grand Challenge 2, pose prediction is well performed for many methodologies, but scoring remains very challenging even when crystal structures are provided.
“According to the results obtained in the Grand Challenge 2, it is clear that the pose prediction task is well performed for many methodologies, but scoring is still a very challenging task, even when the crystal structures are provided (Gaieb et al., 2018).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Gaieb et al., 2018 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationD3R periodically provides pharmaceutical-related benchmark datasets and a Grand Challenge as a blinded community challenge with unpublished data.
“Drug Design Data Resource (D3R10) periodically provide pharmaceutical-related benchmark datasets and a Grand Challenge as a blinded community challenge with unpublished data (Gathiaka et al., 2016).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Gathiaka et al., 2016 source
Not checked in the free run.
May combine two claims.
InterpretationreasoningWrong orientation of hydrogen atoms can lead to high van der Waals energies, underestimation of hydrogen bond interactions, and incorrect electrostatic repulsions between charged/polar groups.
“In such cases, the wrong orientation of hydrogen atoms can lead to high van der Waals energies, underestimation of hydrogen bond interactions, and incorrect electrostatic repulsions between charged/polar groups.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationReliable protein-ligand structures usually comply with criteria: good resolution (2.5 Å or better), fully resolved electron density for the entire ligand and surrounding binding-site residues, and no significant influences from crystal packing on the observed binding mode.
“Reliable protein–ligand structures usually comply these criteria: good resolution (2.5 Å or better), fully resolved electron density for the entire ligand and the surrounding binding-site residues, and without significant influences from crystal packing on the observed binding mode (Cole et al., 2011).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Cole et al., 2011 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningThe broad profile of D3R Grand Challenges makes their datasets one of the more reliable sources to evaluate docking and scoring methods.
“The broad profile of the D3R Grand Challenges, regarding chemical space diversity and affinity data carefully collected, makes their datasets one of the more reliable sources to evaluate docking and scoring methods, providing useful guidelines and best practices for further VS campaigns and methodological improvements.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningProspective evaluations are of substantial importance because blinded predictions simulate real experiments of VS campaigns.
“In addition to the well-known benchmarking sets, prospective evaluations are of substantial importance since the blinded predictions simulate real experiments of VS campaigns.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
InterpretationreasoningScoring functions mainly composed of simple contact-based descriptors (element-element pair counting) emerge to circumvent the complicated preparation required in large datasets for VS.
“In this context, scoring functions mainly composed of simple contact-based descriptors (element–element pair counting) emerge to circumvent the complicated preparation required in large datasets for VS.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationZilian and Sotriffer trained a RF scoring function using the same SFCscore descriptors (SFCscoreRF) and found a much improved model, with R = 0.779 significantly higher than correlations obtained for SFScore linear models.
“As an example, Zilian and Sotriffer (2013) trained a RF scoring function using the same SFCscore descriptors (named SFCscoreRF) and found a much improved model, with R = 0.779 significantly higher than those correlations obtained for the SFScore linear models (Pason and Sotriffer, 2016).”
Quote in the paperSubstantial claim0 of 2 sources checked- cites Zilian and Sotriffer (2013)
Not checked in the free run.
- cites Pason and Sotriffer, 2016
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Premisenone givenScoring functions can be developed using regression methods to reproduce continuous binding constants or classification methods to reproduce binary affinity data.
“Scoring functions can be developed using regression methods to reproduce continuous (e.g., binding constants) or classification methods to reproduce binary affinity data (e.g., active/inactive).”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationLinear scoring functions developed to date have shown moderate correlations (RP ∼ 0.6), whereas nonlinear models achieved significantly better correlations (RP > 0.7) on benchmarking studies.
“Linear scoring functions developed to date have shown moderate correlations (RP ∼ 0.6), whereas nonlinear models achieved significantly better correlations (RP > 0.7) on benchmarking studies (Ashtawy and Mahapatra, 2012; Khamis and Gomaa, 2015; Wang and Zhang, 2017; Wójcikowski et al., 2017).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites Ashtawy and Mahapatra, 2012
Not checked in the free run.
- cites Khamis and Gomaa, 2015
Not checked in the free run.
- cites Wang and Zhang, 2017
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Premisenone givenMany sophisticated machine-learning techniques automatically generate local models for similar training points, being able to classify new instances automatically and use different regression models according to specific properties without explicitly defining classes based on such descriptors.
“Many sophisticated machine-learning techniques automatically generate local models for similar training points (e.g., locally weighted regression), being able to classify the new instances automatically and use different regression models according to specific properties without explicitly defining classes based on such descriptors.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationThe PDBbind refined set follows criteria: only complexes with known Kd or Ki are allowed; no complexes with extremely low (Kd or Ki > 10 mM) or extremely high (Kd or Ki < 1 pM) affinities are accepted; estimated values are rejected.
“the PDBbind refined set follows several criteria concerning the bioactivity manually collected from the original reference (Liu et al., 2015): (i) only complexes with known dissociation constants (Kd) or inhibition constants (Ki) are allowed, (ii) no complexes with extremely low (Kd or Ki > 10 mM) or extremely high (Kd or Ki < 1 pM) affinities are accepted, and (iii) estimated values are rejected”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Liu et al., 2015
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Interpretationnone givenDL methods take advantage of the recent increase in computational power and the ever-expanding availability of structural and binding data.
“Such methods take advantage of the recent increase in computational power and the ever-expanding availability of structural and binding data.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationLarger training sets and describing complexes with more descriptors have a positive impact on the predictive performance of nonlinear models.
“Furthermore, they also observed that larger training sets and describing the complexes with more descriptors have a positive impact in the predictive performance of the nonlinear models.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Li et al. (2014a)
Not checked in the free run.
May combine two claims.
PremisecitationRecently, DL has been applied for pose/affinity prediction and active/inactive detection, exhibiting outstanding performance compared with several well-performing scoring functions developed with both linear and nonlinear approaches.
“Recently, DL has been applied for pose/affinity prediction and active/inactive detection, exhibiting an outstanding performance when compared with several well-performing scoring functions developed with both linear and nonlinear approaches (Wallach et al., 2015; Khamis et al., 2016; Pereira et al., 2016; Ragoza et al., 2017; Jiménez Luna et al., 2018; Nguyen et al., 2018).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites Wallach et al., 2015
Not checked in the free run.
- cites Khamis et al., 2016
Not checked in the free run.
- cites Pereira et al., 2016 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Premisenone givenOver the last decades, many strategies have been implemented in docking programs to consider some degree of flexibility, such as soft potentials and ensemble docking.
“However, over the last decades, many strategies have been implemented in docking programs to consider some degree of flexibility in the targeted, such as soft potentials and ensemble docking.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationcitationDL is considered a promising approach to diverse drug discovery projects guided by successes in image and speech recognition problems.
“DL is considered as a promising approach to diverse drug discovery projects guided by the successes obtained in image and speech recognition problems (Zhang L. et al., 2017).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Zhang L. et al., 2017
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationBodnarchuk published an extensive review of water-placement methods helpful for locating conserved water molecules within the protein binding site to be considered explicitly during docking simulation.
“Recently, Bodnarchuk (2016) published an extensive review of water-placement methods helpful for locating conserved water molecules within the protein binding site to be considered explicitly during the docking simulation.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Bodnarchuk (2016) source
Not checked in the free run.
May combine two claims.
Interpretationnone givenSolvent mapping approaches only calculate the solvent contribution to the free energy and must be combined with other methods to be used for lead optimization or VS.
“However, these approaches only calculate the solvent contribution to the free energy and must be combined with other methods to be used for lead optimization or VS.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationCovalent inhibitors can be divided into reversible and irreversible categories according to whether inhibition is reversible or irreversible.
“Covalent inhibitors can further be divided into two different categories according to whether inhibition is reversible or irreversible (Tuley and Fast, 2018).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Tuley and Fast, 2018
Not checked in the free run.
May combine two claims.
PremisecitationMany solvent mapping methods evaluated on real drug design studies showed that solvent mapping methods could be important to help ligand optimization and to correctly rank compounds to assist synthetic prioritization.
“Many solvent mapping methods were evaluated on real drug design studies in a recent paper (Bucher et al., 2018), showing that solvent mapping methods could be important to help ligand optimization and to correctly rank compounds to assist synthetic prioritization.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Bucher et al., 2018 source
Not checked in the free run.
May combine two claims.
Interpretationnone givenWhereas high computational cost related with sampling protein conformations and docking large compound libraries can be overcome with high-performance computing platforms, weighing such conformations and integrating them with scoring functions remains a hindrance for accurate estimation of binding affinities on flexible systems.
“Whereas the high computational cost related with sampling protein conformations and docking large compound libraries can be overcome with the use of high-performance computing platforms, weighing such conformations and integrating them with the scoring functions remains a hindrance for accurate estimation of binding affinities on flexible systems.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationIn AutoDock4, explicit water molecules of the first hydration shell are represented as uncharged spheres directly attached to the ligand, whereas a hydration force field accounting for entropic and enthalpic contributions automatically predicts their potential in mediating protein-ligand interactions.
“In AutoDock4, explicit water molecules of the first hydration shell as represented as uncharged spheres directly attached to the ligand, whereas a hydration force field accounting for the entropic and enthalpic contributions, automatically predicts their potential in mediating protein–ligand interactions (Forli and Olson, 2012).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Forli and Olson, 2012 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationA water-mapping protocol based on post-trajectory analysis of explicit solvent MD uses inhomogeneous solvation theory and tries to predict the free energy cost of moving a water molecule from a protein hydration site into the bulk solvent.
“An interesting approach is the use of a water-mapping protocol based on the post trajectory analysis of explicit solvent MD. This analysis is based on the inhomogeneous solvation theory and tries to predict the free energy cost of moving a water molecule from a protein hydration site into the bulk solvent (Yang et al., 2013).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Yang et al., 2013 source
Not checked in the free run.
May combine two claims.
InterpretationreasoningThese advances were essential to overcome the bottleneck of high computational cost and are allowing the increasing use of QM methods.
“These advances were essential to overcome the bottleneck of the high computational cost and are allowing the increasing use of QM methods in the”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationcitationDevelopment of covalent-docking methodologies is very important due to potential advantages of covalent inhibitors, including sustained duration of action, increased ligand efficiency, ability to inhibit shallow binding sites previously categorized as undruggable, and increased ability to overcome resistant mutations.
“The development of covalent-docking methodologies capable of dealing with such type of inhibition is very important due to the potential advantages associated with covalent inhibitors (De Cesco et al., 2017), including (i) sustained duration of action leading to less frequent dosing, (ii) increased ligand efficiency, (iii) ability to inhibit targets with shallow binding sites previously”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites De Cesco et al., 2017 source
Not checked in the free run.
May combine two claims.
PremisecitationIn the last 10 years, important advances were made in computing hardware, quantum algorithms to compute molecular wave functions, more reliable semi-empirical quantum methods, and new hybrid QM/MM methods.
“In the last 10 years, important advances were made in computing hardware (e.g., Graphics Processing Units – GPUs), in the development of quantum algorithms to compute molecular wave functions (Dixon and Merz, 1997; Birgin et al., 2013), the development of more reliable semi-empirical quantum methods (Christensen et al., 2016; Yilmazer and Korth, 2016), and development of new hybrid QM/MM methods (”
Quote in the paperNot a substantial claim0 of 3 sources checkedMay combine two claims.
InterpretationreasoningThese non-classical interactions/effects are beyond the limits of classical methods and represent a significant challenge to the development of scoring functions for computational drug design.
“These non-classical interactions/effects are beyond the limits of classical methods and represent a significant challenge to the development of scoring functions to be used in computational drug design experiments.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Interpretationnone givenThe development of docking methodologies to predict poses and binding affinities of ligands that bind covalently to the receptor is a challenging task.
“The development of docking methodologies to predict poses and binding affinities of ligands that bind covalently to the receptor is a challenging task.”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationreasoningMetal ion interactions are essential when dealing with metalloproteins and, due to large changes in electronic structure under ligand binding, are also a great challenge.
“In particular, metal ions interactions are essential when dealing with metalloproteins and, due to the large changes in the electronic structure under ligand binding, are also a great challenge.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationDue to increasing interest in covalent drugs, many non-covalent docking programs have developed covalent versions and some new docking programs focused on covalent ligands have been developed.
“Due to the increasing interest in covalent drugs, many non-covalent docking programs have developed covalent versions and some new docking programs focused on covalent ligands have been developed (Kumalo et al., 2015; Awoonor-Williams et al., 2017; De Cesco et al., 2017).”
Quote in the paperNot a substantial claim0 of 3 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
InterpretationcitationTwo critical aspects in the future development of covalent scoring functions are the capacity to predict the kinetics of ligand binding and the intrinsic reactivity of electrophilic and nucleophilic pairs of atoms.
“Two critical aspects in the future development of covalent scoring functions are the capacity to predict the kinetics of ligand binding (e.g., residence times) and the intrinsic reactivity of electrophilic and nucleophilic pairs of atoms (De Cesco et al., 2017).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites De Cesco et al., 2017 source
Not checked in the free run.
May combine two claims.
PremisecitationAdvances in QM methods were essential to overcome the bottleneck of high computational cost and are allowing increasing use of QM methods in the prediction of protein-ligand binding affinities.
“These advances were essential to overcome the bottleneck of the high computational cost and are allowing the increasing use of QM methods in the prediction of protein–ligand binding affinities (Crespo et al., 2017).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Crespo et al., 2017 source
Not checked in the free run.
May combine two claims.
InterpretationcitationThe use of quantum mechanical methods can improve the description of protein-ligand interactions and, in principle, could provide a more accurate binding affinity.
“The use of quantum mechanical methods can improve the description of protein–ligand interactions and, in principle, could provide a more accurate binding affinity (Raha and Merz, 2005; Chaskar et al., 2017; Crespo et al., 2017; Cavasotto et al., 2018).”
Quote in the paperNot a substantial claim0 of 3 sources checkedMay combine two claims.
PremisecitationThe XBScoreQM scoring function achieved good performance on both pose and affinity prediction when compared against 12 diverse scoring functions and increased predictive capacity for protein-ligand complexes with X-bond interactions.
“The new scoring function achieved good performances on both pose and affinity prediction when compared against 12 diverse scoring functions, and increase predictive capacity to deal with protein–ligand complexes with X-bond interactions.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Yang et al., 2015 source
Not checked in the free run.
May combine two claims.
InterpretationcitationConsensus scoring excels over individual scoring functions in docking-based VS and is less sensitive to protein target variation.
“In addition, they compared their machine-learning-based CS with individual scoring functions and traditional CS schemed, confirming that CS excel individual scoring functions performances in docking-based VS, being less sensitive to protein target variation.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Ericksen et al., 2017 source
Not checked in the free run.
May combine two claims.
PremisecitationEricksen et al. (2017) developed machine-learning consensus scoring using discrete mixture models and gradient boosting to combine scores from eight docking programs and obtained improved performance over individual scoring functions on 21 targets from the DUD-E dataset.
“For example, Ericksen et al. (2017) developed machine-learning CS using discrete mixture models and gradient boosting to combine the scores from eight docking programs and obtained improved performances than individual scoring functions on 21 targets from DUD-E dataset.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Ericksen et al., 2017 source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based).
Interpretationnone givenIn recent years, the increasing availability of protein-ligand structures with measured binding affinities and datasets containing active, decoy, and true inactive compounds are boosting the use of sophisticated machine-learning techniques to obtain better performing scoring functions.
“In recent years, the increasing availability of protein–ligand structures with measured binding affinities and data sets containing active, decoy, and true inactive compounds are boosting the use of sophisticated machine-learning techniques to obtain better performing scoring functions.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S2 (checkable), S4 (based).
Interpretationnone givenMethodological advances will be dependent on the size and quality of the available datasets for training and benchmarking.
“Methodological advances will be dependent to the size and quality of the available datasets for training and benchmarking”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationTarget-specific scoring functions trained with RF only performed slightly better than generic models, with two-thirds of them increasing EF1% less than 10%.
“Still, in the recent work of Wójcikowski et al. (2017), the target-specific scoring functions trained with RF only performed slightly better than generic models, with two-third of them increasing the EF1% less than 10%.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wójcikowski et al., 2017 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
PremisereasoningTo overcome the lack of a sufficient amount of experimental structures, protein-ligand conformations used for training target-specific scoring functions are commonly obtained from docking experiments.
“To overcome the lack of a sufficient amount of experimental structures, protein–ligand conformations used for training target-specific scoring functions are commonly obtained from docking experiments.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationInhibitors of protein-protein interactions interact in uncommon binding cavities characterized by higher hydrophobicity, aromaticity, and molecular weight compared to enzyme inhibitors, and usually within flatter, larger, and more hydrophobic binding sites than enzyme catalytic sites.
“which interact in uncommon binding cavities characterized by higher hydrophobicity, aromaticity, and molecular weight compared to enzyme inhibitors, as usually interacting within flatter, larger, and more hydrophobic binding sites than the enzyme catalytic sites (Morelli et al., 2011; Kuenemann et al., 2014).”
Quote in the paperNot a substantial claim0 of 2 sources checkedProbably not a substantial claim: may not pass S2 (checkable), S4 (based).
Interpretationnone givenReliable predictions of ligand binding affinity remain a big challenge.
“Reliable predictions of ligand binding affinity remain a big challenge”
Quote in the paperNot a substantial claim- Rests on an assumption: The development of accurate empirical scoring functions to predict protein–ligand binding affinities is a key aspect in SBDD.
May combine two claims.
PremisecitationTailored scoring functions are more beneficial for protein targets with fewer active compounds than for those containing more actives, where target-specific scoring functions exhibit similar performance to the generic model.
“As an intriguing result, they found that tailored scoring functions are more beneficial for the protein targets with less active compounds than the others containing more actives, where the target-specific scoring functions exhibit similar performances to the generic model.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites Wójcikowski et al., 2017 source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based).
Interpretationnone givenThe requirement of a large training set to derive a robust scoring function might become a significant hindrance and source of inaccuracy.
“Despite encouraging results obtained for target-specific scoring functions, it is important to highlight that the requirement of a large training set to derive a robust scoring function might become a significant hindrance and source of inaccuracy.”
Quote in the paperNot a substantial claim- Rests on an assumption: Target-specific scoring functions yield significant improvements in docking and VS accuracies compared to non-specific models when trained on target-specific protein–ligand complexes.
Probably not a substantial claim: may not pass S2 (checkable).
InterpretationreasoningGreat care will be necessary to avoid artificial performances due to the increased capacity of nonlinear methods to capture bias present in the training data.
“great care will be necessary to avoid artificial performances due to the increased capacity of these nonlinear methods to capture bias present in the training data.”
Quote in the paperNot a substantial claim
The paper relies on this without support.
Empirical scoring functions are widely used for pose and affinity prediction.
“Empirical scoring functions are widely used for pose and affinity prediction.”
The paper relies on this without support.
There is no universal scoring function with significant reliability for all molecular systems.
“Although there is no universal scoring function with significant reliability for all molecular systems, some important strategies were explored.”
The paper relies on this without support.
The methodology behind empirical scoring functions is fast enough to be used in large-scale structure-based VS and de novo design studies.
“First, the methodology behind this type of scoring function could be fast enough to be used in large-scale structure-based VS and de novo design studies.”
The paper relies on this without support.
The use of modern sophisticated machine-learning techniques and the increasing availability of protein–ligand structures and measured binding affinity data will increase the accuracy of empirical scoring functions enough to be useful in computer-aided SBDD experiments.
“Secondly, the use of modern sophisticated machine-learning techniques and the increasing availability of protein–ligand structures and measured binding affinity data could increase considerably the accuracy of empirical scoring functions to be useful in computer-aided SBDD experiments.”
The authors state this as an assumption.
All discussion in this review assumes that the inhibitors being dealt with are non-covalent.
“All the discussion made in this review assumes that we are dealing with non-covalent inhibitors.”
The paper relies on this without support.
Target-specific scoring functions yield significant improvements in docking and VS accuracies compared to non-specific models when trained on target-specific protein–ligand complexes.
“Significant improvements in docking and VS accuracies are reported when employing target-specific scoring functions rather than non-specific models, using as training datasets protein–ligand complexes comprising specific molecular targets instead of a general dataset.”
The paper relies on this without support.
The development of accurate empirical scoring functions to predict protein–ligand binding affinities is a key aspect in SBDD.
“The development of accurate empirical scoring functions to predict protein–ligand binding affinities is a key aspect in SBDD.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The Rosetta all-atom scoring function often has severe limitations.
“However, there are many limitations to these techniques, as they are dependent on databases of known structure, are computationally intensive (such as involving significant sampling), as well as being dependent on classical scoring functions, such as the Rosetta all-atom scoring function, which often have severe limitations [95].”
From Computational and artificial intelligence-based methods for antibody development (2023) · cites this paper as [95]
Classical antibody design techniques are dependent on databases of known structure.
“However, there are many limitations to these techniques, as they are dependent on databases of known structure, are computationally intensive (such as involving significant sampling), as well as being dependent on classical scoring functions, such as the Rosetta all-atom scoring function, which often have severe limitations [95].”
From Computational and artificial intelligence-based methods for antibody development (2023) · cites this paper as [95]
Classical antibody design techniques are dependent on classical scoring functions.
“However, there are many limitations to these techniques, as they are dependent on databases of known structure, are computationally intensive (such as involving significant sampling), as well as being dependent on classical scoring functions, such as the Rosetta all-atom scoring function, which often have severe limitations [95].”
From Computational and artificial intelligence-based methods for antibody development (2023) · cites this paper as [95]
Classical antibody design techniques are computationally intensive, involving significant sampling.
“However, there are many limitations to these techniques, as they are dependent on databases of known structure, are computationally intensive (such as involving significant sampling), as well as being dependent on classical scoring functions, such as the Rosetta all-atom scoring function, which often have severe limitations [95].”
From Computational and artificial intelligence-based methods for antibody development (2023) · cites this paper as [95]
Cites
Berman, H. M. (2000) · Nucleic Acids Research · doi:10.1093/nar/28.1.235 · Cited by 40,196 · see citing works
Trott, O., & Olson, A. J. (2010) · Journal of Computational Chemistry · doi:10.1002/jcc.21334 · Cited by 38,412 · see citing works
Friesner, R. A., Banks, J. L., Murphy, R. B., Halgren, T. A., Klicic, J. J., Mainz, D. T., Repasky, M. P., Knoll, E. H., Shelley, M., Perry, J. K., et al. (2004) · Journal of Medicinal Chemistry · doi:10.1021/jm0306430 · Cited by 10,057 · see citing works
Friesner, R. A., Murphy, R. B., Repasky, M. P., Frye, L. L., Greenwood, J. R., Halgren, T. A., Sanschagrin, P. C., & Mainz, D. T. (2006) · Journal of Medicinal Chemistry · doi:10.1021/jm051256o · Cited by 6,964 · see citing works
Development and validation of a genetic algorithm for flexible docking 1 1Edited by F. E. Cohen
Jones, G., Willett, P., Glen, R. C., Leach, A. R., & Taylor, R. (1997) · Journal of Molecular Biology · doi:10.1006/jmbi.1996.0897 · Cited by 6,763 · see citing works
Sastry, G. M., Adzhigirey, M., Day, T., Annabhimoju, R., & Sherman, W. (2013) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-013-9644-8 · Cited by 6,130 · see citing works
Merck molecular force field. I. Basis, form, scope, parameterization, and performance of MMFF94
Halgren, T. A. (1996) · Journal of Computational Chemistry · doi:10.1002/(sici)1096-987x(199604)17:5/6 · Cited by 5,321 · see citing works
Pattern Recognition and Machine Learning
Neal, R. M. (2007) · Technometrics · doi:10.1198/tech.2007.s518 · Cited by 4,629 · see citing works
Golbraikh, A., & Tropsha, A. (2002) · Journal of Molecular Graphics and Modelling · doi:10.1016/s1093-3263(01)00123-1 · Cited by 3,735 · see citing works
Semianalytical treatment of solvation for molecular mechanics and dynamics
Still, W. C., Tempczyk, A., Hawley, R. C., & Hendrickson, T. F. (1990) · Journal of the American Chemical Society · doi:10.1021/ja00172a038 · Cited by 3,653 · see citing works
Innovation in the pharmaceutical industry: New estimates of R&D costs
DiMasi, J. A., Grabowski, H. G., & Hansen, R. W. (2016) · Journal of Health Economics · doi:10.1016/j.jhealeco.2016.01.012 · Cited by 3,359 · see citing works
A Fast Flexible Docking Method using an Incremental Construction Algorithm
Rarey, M., Kramer, B., Lengauer, T., & Klebe, G. (1996) · Journal of Molecular Biology · doi:10.1006/jmbi.1996.0477 · Cited by 2,836 · see citing works
The HADDOCK2.2 Web Server: User-Friendly Integrative Modeling of Biomolecular Complexes
van Zundert, G. C. P., Rodrigues, J. P. G. L. M., Trellet, M., Schmitz, C., Kastritis, P. L., Karaca, E., Melquiond, A. S. J., van Dijk, M., de Vries, S. J., & Bonvin, A. M. J. J. (2016) · Journal of Molecular Biology · doi:10.1016/j.jmb.2015.09.014 · Cited by 2,785 · see citing works
Molecular Docking and Structure-Based Drug Design Strategies
Ferreira, L. L. G., dos Santos, R. N., Oliva, G., & Andricopulo, A. D. (2015) · Molecules · doi:10.3390/molecules200713384 · Cited by 2,642 · see citing works
Hawkins, D. M. (2004) · Journal of Chemical Information and Computer Sciences · doi:10.1021/ci0342472 · Cited by 2,522 · see citing works
Directory of useful decoys, enhanced (DUD-E): better ligands and decoys for better benchmarking.
Mysinger, M. M., Carchia, M., Irwin, J. J., & Shoichet, B. K. (2012) · Journal of Medicinal Chemistry · doi:10.1021/jm300687e · Cited by 2,412 · see citing works
On the determination of molecular fields. —II. From the equation of state of a gas
Jones, J. E. (1924) · Proceedings of the Royal Society of London Series A Containing Papers of a Mathematical and Physical Character · doi:10.1098/rspa.1924.0082 · Cited by 2,403 · see citing works
Accurate Calculation of Hydration Free Energies Using Macroscopic Solvent Models
Sitkoff, D. F., Sharp, K. A., & Honig, B. H. (1994) · The Journal of Physical Chemistry · doi:10.1021/j100058a043 · Cited by 2,160 · see citing works
Definition of the halogen bond (IUPAC Recommendations 2013).
Desiraju, G. R., Ho, P. S., Kloo, L., Legon, A. C., Marquardt, R., Metrangolo, P., Politzer, P., Resnati, G., & Rissanen, K. (2013) · Pure and Applied Chemistry · doi:10.1351/pac-rec-12-05-10 · Cited by 1,957 · see citing works
Linear Model Selection by Cross-validation
Shao, J. (1993) · Journal of the American Statistical Association · doi:10.1080/01621459.1993.10476299 · Cited by 1,790 · see citing works
Eldridge, M., Murray, C. W., Auton, T. R., Paolini, G. V., & Mee, R. P. (1997) · Journal of Computer-Aided Molecular Design · doi:10.1023/a:1007996124545 · Cited by 1,692 · see citing works
Abagyan, R., Totrov, M., & Kuznetsov, D. (1994) · Journal of Computational Chemistry · doi:10.1002/jcc.540150503 · Cited by 1,667 · see citing works
Software for molecular docking: a review
Pagadala, N. S., Syed, K., & Tuszynski, J. (2017) · Biophysical Reviews · doi:10.1007/s12551-016-0247-1 · Cited by 1,522 · see citing works
Benchmarking sets for molecular docking.
Huang, N., Shoichet, B. K., & Irwin, J. J. (2006) · Journal of Medicinal Chemistry · doi:10.1021/jm0608356 · Cited by 1,318 · see citing works
Wang, R., Lai, L., & Wang, S. (2002) · Journal of Computer-Aided Molecular Design · doi:10.1023/a:1016357811882 · Cited by 1,185 · see citing works
WATER MEDIATION IN PROTEIN FOLDING AND MOLECULAR RECOGNITION
Levy, Y., & Onuchic, J. N. (2006) · Annual Review of Biophysics and Biomolecular Structure · doi:10.1146/annurev.biophys.35.040405.102134 · Cited by 1,051 · see citing works
B�hm, H.-J. (1994) · Journal of Computer-Aided Molecular Design · doi:10.1007/bf00126743 · Cited by 1,013 · see citing works
Ballester, P. J., & Mitchell, J. B. O. (2010) · Bioinformatics · doi:10.1093/bioinformatics/btq112 · Cited by 988 · see citing works
Calculation of Protein-Ligand Binding Affinities
Gilson, M. K., & Zhou, H. (2007) · Annual Review of Biophysics and Biomolecular Structure · doi:10.1146/annurev.biophys.36.040306.132550 · Cited by 966 · see citing works
Qiu, D., Shenkin, P. S., Hollinger, F. P., & Still, W. C. (1997) · The Journal of Physical Chemistry A · doi:10.1021/jp961992r · Cited by 962 · see citing works
Computer-based de novo design of drug-like molecules
Schneider, G., & Fechner, U. (2005) · Nature Reviews Drug Discovery · doi:10.1038/nrd1799 · Cited by 956 · see citing works
Automated docking with grid‐based energy evaluation
Meng, E. C., Shoichet, B. K., & Kuntz, I. D. (1992) · Journal of Computational Chemistry · doi:10.1002/jcc.540130412 · Cited by 956 · see citing works
K DEEP : Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks
Jiménez-Luna, J., Škalič, M., Martínez-Rosell, G., & Fabritiis, G. D. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00650 · Cited by 923 · see citing works
Folding of chymotrypsin inhibitor 2. 1. Evidence for a two-state transition
Jackson, S., & Fersht, A. R. (1991) · Biochemistry · doi:10.1021/bi00107a010 · Cited by 888 · see citing works
Jones, J. E. (1924) · Proceedings of the Royal Society of London Series A Containing Papers of a Mathematical and Physical Character · doi:10.1098/rspa.1924.0081 · Cited by 864 · see citing works
Comparative evaluation of 11 scoring functions for molecular docking.
Wang, R., Lu, Y., & Wang, S. (2003) · Journal of Medicinal Chemistry · doi:10.1021/jm0203783 · Cited by 849 · see citing works
Gohlke, H., & Klebe, G. (2002) · Angewandte Chemie International Edition · doi:10.1002/1521-3773(20020802)41:15 · Cited by 843 · see citing works
The computer program LUDI: A new method for the de novo design of enzyme inhibitors
B�hm, H.-J. (1992) · Journal of Computer-Aided Molecular Design · doi:10.1007/bf00124387 · Cited by 840 · see citing works
Protein–Ligand Scoring with Convolutional Neural Networks
Ragoza, M., Hochuli, J. E., Idrobo, E., Sunseri, J., & Koes, D. R. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.6b00740 · Cited by 835 · see citing works
A surprising simplicity to protein folding
Baker, D. A. (2000) · Nature · doi:10.1038/35011000 · Cited by 778 · see citing works
From machine learning to deep learning: progress in machine intelligence for rational drug discovery
Zhang, L., Tan, J., Han, D., & Zhu, H. (2017) · Drug Discovery Today · doi:10.1016/j.drudis.2017.08.010 · Cited by 760 · see citing works
Bissantz, C., Folkers, G., & Rognan, D. (2000) · Journal of Medicinal Chemistry · doi:10.1021/jm001044l · Cited by 740 · see citing works
Sun, H., Li, Y., Tian, S., Xu, L., & Hou, T. (2014) · Phys. Chem. Chem. Phys. · doi:10.1039/c4cp01388c · Cited by 738 · see citing works
DOCK 6: Impact of New Features and Current Docking Performance
Allen, W. J., Balius, T. E., Mukherjee, S., Brozell, S., Moustakas, D., Lang, P. T., Case, D., Kuntz, I., & Rizzo, R. C. (2015) · Journal of Computational Chemistry · doi:10.1002/jcc.23905 · Cited by 731 · see citing works
The hydrophobic effect: Formation of micelles and biological membranes
Gruen, D. W. R., & Haydon, D. A. (1980) · Journal of Colloid and Interface Science · doi:10.1016/0021-9797(80)90338-0 · Cited by 713 · see citing works
AutoDockFR: Advances in Protein-Ligand Docking with Explicitly Specified Binding Site Flexibility
Ravindranath, P. A., Forli, S., Goodsell, D. S., Olson, A. J., & Sanner, M. F. (2015) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1004586 · Cited by 700 · see citing works
Charifson, P. S., Corkery, J. J., Murcko, M. A., & Walters, W. P. (1999) · Journal of Medicinal Chemistry · doi:10.1021/jm990352k · Cited by 698 · see citing works
Wallach, I., Dzamba, M., & Heifets, A. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1510.02855 · Cited by 698 · see citing works
PDB-wide collection of binding data: current status of the PDBbind database
Liu, Z., Li, Y., Han, L., Li, J., Liu, J., Zhao, Z., Nie, W., Liu, Y., & Wang, R. (2014) · Bioinformatics · doi:10.1093/bioinformatics/btu626 · Cited by 650 · see citing works
Receptor–ligand molecular docking
Guedes, I. A., de Magalhães, C. S., & Dardenne, L. E. (2013) · Biophysical Reviews · doi:10.1007/s12551-013-0130-2 · Cited by 611 · see citing works
Self-citationJain, A. N. (1996) · Journal of Computer-Aided Molecular Design · doi:10.1007/bf00124474 · Cited by 609 · see citing works
Comparative Assessment of Scoring Functions on a Diverse Test Set
Cheng, T., Li, X., Li, Y., Liu, Z., & Wang, R. (2009) · Journal of Chemical Information and Modeling · doi:10.1021/ci9000053 · Cited by 514 · see citing works
Macroscopic models of aqueous solutions: biological and chemical applications
Honig, B. H., Sharp, K. A., & Yang, A. (1993) · The Journal of Physical Chemistry · doi:10.1021/j100108a002 · Cited by 514 · see citing works
Moitessier, N., Englebienne, P., Lee, D., Lawandi, J., & Corbeil, C. R. (2008) · British Journal of Pharmacology · doi:10.1038/sj.bjp.0707515 · Cited by 514 · see citing works
Huang, S.-Y., Grinter, S. Z., & Zou, X. (2010) · Physical Chemistry Chemical Physics · doi:10.1039/c0cp00151a · Cited by 508 · cited 1× · see citing works
“In this context, scoring functions emerge as a straightforward and fast strategy despite limited accuracy, remaining as the main alternative to be applied in VS experiments (Huang et al., 2010).”
MethodForging the basis for developing protein-ligand interaction scoring functions.
Liu, Z., Su, M., Han, L., Liu, J., Yang, Q., Li, Y., & Wang, R. (2017) · Accounts of Chemical Research · doi:10.1021/acs.accounts.6b00491 · Cited by 493 · see citing works
Do enthalpy and entropy distinguish first in class from best in class?
Freire, E. (2008) · Drug Discovery Today · doi:10.1016/j.drudis.2008.07.005 · Cited by 491 · see citing works
Variability in docking success rates due to dataset preparation.
Corbeil, C. R., Williams, C. I., & Labute, P. (2012) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-012-9570-1 · Cited by 469 · see citing works
Christensen, A. S., Kubař, T., Cui, Q., & Elstner, M. (2016) · Chemical Reviews · doi:10.1021/acs.chemrev.5b00584 · Cited by 446 · see citing works
Docking Covalent Inhibitors: A Parameter Free Approach To Pose Prediction and Scoring
Zhu, K., Borrelli, K., Greenwood, J., Day, T., Abel, R., Farid, R., & Harder, E. D. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci500118s · Cited by 445 · see citing works
Ligand configurational entropy and protein binding
Chang, C. A., Chen, W., & Gilson, M. K. (2007) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.0610494104 · Cited by 439 · see citing works
Boosting docking-based virtual screening with deep learning.
Pereira, J. C., Caffarena, E. R., & dos Santos, C. N. (2016) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.6b00355 · Cited by 395 · see citing works
Modeling water molecules in protein-ligand docking using GOLD.
Verdonk, M. L., Chessari, G., Cole, J. C., Hartshorn, M. J., Murray, C. W., Nissink, J. W. M., Taylor, R. D., & Taylor, R. (2005) · Journal of Medicinal Chemistry · doi:10.1021/jm050543p · Cited by 393 · see citing works
Li, Y., Han, L., Liu, Z., & Wang, R. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci500081m · Cited by 384 · see citing works
LigScore: a novel scoring function for predicting binding affinities
Krammer, A., Kirchhoff, P. D., Jiang, X., Venkatachalam, C. M., & Waldman, M. (2004) · Journal of Molecular Graphics and Modelling · doi:10.1016/j.jmgm.2004.11.007 · Cited by 383 · see citing works
Virtual Screening Using Protein−Ligand Docking: Avoiding Artificial Enrichment
Verdonk, M. L., Berdini, V., Hartshorn, M. J., Mooij, W. T. M., Murray, C. W., Taylor, R. D., & Watson, P. (2004) · Journal of Chemical Information and Computer Sciences · doi:10.1021/ci034289q · Cited by 379 · see citing works
Performance of machine-learning scoring functions in structure-based virtual screening.
Wójcikowski, M., Ballester, P. J., & Siedlecki, P. (2017) · Scientific Reports · doi:10.1038/srep46710 · Cited by 364 · see citing works
Ain, Q. U., Aleksandrova, A., Roessler, F. D., & Ballester, P. J. (2015) · WIREs Computational Molecular Science · doi:10.1002/wcms.1225 · Cited by 357 · see citing works
Recommendations for evaluation of computational methods
Jain, A. N., & Nicholls, A. O. (2008) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-008-9196-5 · Cited by 350 · see citing works
Additivity principles in biochemistry.
Dill, K. A. (1997) · Journal of Biological Chemistry · doi:10.1074/jbc.272.2.701 · Cited by 347 · see citing works
Velec, H. F. G., Gohlke, H., & Klebe, G. (2005) · Journal of Medicinal Chemistry · doi:10.1021/jm050436v · Cited by 346 · see citing works
Latest developments in molecular docking: 2010–2011 in review
Yuriev, E., & Ramsland, P. A. (2013) · Journal of Molecular Recognition · doi:10.1002/jmr.2266 · Cited by 344 · see citing works
Improving scoring-docking-screening powers of protein-ligand scoring functions using random forest
Wang, C., & Zhang, Y. (2016) · Journal of Computational Chemistry · doi:10.1002/jcc.24667 · Cited by 333 · see citing works
Linear Model Selection by Cross-Validation
Shao, J. (1993) · Journal of the American Statistical Association · doi:10.2307/2290328 · Cited by 327 · see citing works
Santos-Martins, D., Forli, S., Ramos, M. J., & Olson, A. J. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci500209e · Cited by 318 · cited 1× · see citing works
“AutoDock4Zn has implemented a specific force-field-based potential for the zinc ion to consider both geometric and energetic components of the metal–ligand interaction, achieving better performance for pose prediction in redocking experiments (Santos-Martins et al., 2014).”
MethodCovalent inhibitors design and discovery.
De Cesco, S., Kurian, J., Dufresne, C., Mittermaier, A. K., & Moitessier, N. (2017) · European Journal of Medicinal Chemistry · doi:10.1016/j.ejmech.2017.06.019 · Cited by 315 · see citing works
How Does Consensus Scoring Work for Virtual Library Screening? An Idealized Computer Experiment
Wang, R., & Wang, S. (2001) · Journal of Chemical Information and Computer Sciences · doi:10.1021/ci010025x · Cited by 312 · see citing works
Muegge, I. (2005) · Journal of Medicinal Chemistry · doi:10.1021/jm050038s · Cited by 306 · see citing works
Interpreting steep dose-response curves in early inhibitor discovery.
Shoichet, B. K. (2006) · Journal of Medicinal Chemistry · doi:10.1021/jm061103g · Cited by 304 · see citing works
Classification of Current Scoring Functions
Liu, J., & Wang, R. (2015) · Journal of Chemical Information and Modeling · doi:10.1021/ci500731a · Cited by 302 · see citing works
Ligand-Binding Affinity Estimates Supported by Quantum-Mechanical Methods
Ryde, U., & Söderhjelm, P. (2016) · Chemical Reviews · doi:10.1021/acs.chemrev.5b00630 · Cited by 301 · see citing works
MTiOpenScreen: a web server for structure-based virtual screening
Labbé, C. M., Rey, J., Lagorce, D., Vavruša, M., Becot, J., Spérandio, O., Villoutreix, B. O., Tufféry, P., & Miteva, M. A. (2015) · Nucleic Acids Research · doi:10.1093/nar/gkv306 · Cited by 297 · see citing works
A Force Field with Discrete Displaceable Waters and Desolvation Entropy for Hydrated Ligand Docking
Forli, S., & Olson, A. J. (2012) · Journal of Medicinal Chemistry · doi:10.1021/jm2005145 · Cited by 290 · see citing works
Automated Docking Screens: A Feasibility Study
Irwin, J. J., Shoichet, B. K., Mysinger, M. M., Huang, N., Colizzi, F., Wassam, P., & Cao, Y. (2009) · Journal of Medicinal Chemistry · doi:10.1021/jm9006966 · Cited by 288 · see citing works
Identification of Protein–Protein Interaction Sites from Docking Energy Landscapes
Fernández‐Recio, J., Totrov, M. M., & Abagyan, R. A. (2003) · Journal of Molecular Biology · doi:10.1016/j.jmb.2003.10.069 · Cited by 274 · see citing works
Covalent docking of large libraries for the discovery of chemical probes
London, N., Miller, R. M., Krishnan, S., Uchida, K., Irwin, J. J., Eidam, O., Gibold, L., Cimermančič, P., Bonnet, R., Shoichet, B. K., et al. (2014) · Nature Chemical Biology · doi:10.1038/nchembio.1666 · Cited by 274 · see citing works
Grinter, S. Z., & Zou, X. (2014) · Molecules · doi:10.3390/molecules190710150 · Cited by 264 · see citing works
Improvements, trends, and new ideas in molecular docking: 2012–2013 in review
Yuriev, E., Holien, J. K., & Ramsland, P. A. (2015) · Journal of Molecular Recognition · doi:10.1002/jmr.2471 · Cited by 264 · see citing works
Scoring functions for protein-ligand docking.
Bentham Science Publisher, B. S. P. (2006) · Current Protein & Peptide Science · doi:10.2174/138920306778559395 · Cited by 258 · cited 1× · see citing works
“Given the difficulty in modeling entropic effects for (cid:49) G bind , scoring functions generally neglect their contributions or adopt simplified algorithms to approximate entropies in a straightforward manner (Jain, 2006).”
BackgroundLi, H., Leung, K., Wong, M., & Ballester, P. J. (2015) · Molecular Informatics · doi:10.1002/minf.201400132 · Cited by 252 · see citing works
SCORE: A New Empirical Method for Estimating the Binding Affinity of a Protein-Ligand Complex
Wang, R., Liu, L., Lai, L., & Youqi, T. (1998) · Journal of Molecular Modeling · doi:10.1007/s008940050096 · Cited by 249 · see citing works
Covalent docking using autodock: two-point attractor and flexible side chain methods.
Bianco, G., Forli, S., Goodsell, D. S., & Olson, A. J. (2016) · Protein Science · doi:10.1002/pro.2733 · Cited by 246 · see citing works
Consensus scoring criteria for improving enrichment in virtual screening.
Yang, J.-M., Chen, Y.-F., Shen, T.-W., Kristal, B. S., & Hsu, D. F. (2005) · Journal of Chemical Information and Modeling · doi:10.1021/ci050034w · Cited by 241 · see citing works
Soft Docking and Multiple Receptor Conformations in Virtual Screening
Ferrari, A. M., Wei, B. Q., Costantino, L., & Shoichet, B. K. (2004) · Journal of Medicinal Chemistry · doi:10.1021/jm049756p · Cited by 239 · see citing works
A model binding site for testing scoring functions in molecular docking.
Wei, B. Q., Baase, W. A., Weaver, L. H., Matthews, B. W., & Shoichet, B. K. (2002) · Journal of Molecular Biology · doi:10.1016/s0022-2836(02)00777-5 · Cited by 237 · see citing works
Martin, Y. C. (2009) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-009-9303-2 · Cited by 237 · see citing works
The Experimental Uncertainty of Heterogeneous Public Ki Data
Krämer, C., Kalliokoski, T., Gedeck, P., & Vulpetti, A. (2012) · Journal of Medicinal Chemistry · doi:10.1021/jm300131x · Cited by 234 · see citing works
NAMD goes quantum: an integrative suite for hybrid simulations
Melo, M. C. R., Bernardi, R. C., Rudack, T., et al. (2018) · Nature Methods · doi:10.1038/nmeth.4638 · Cited by 230 · see citing works
New drugs cost US$2.6 billion to develop
Mullard, A. (2014) · Nature Reviews Drug Discovery · doi:10.1038/nrd4507 · Cited by 228 · see citing works
Kinase inhibitors and the case for CH…O hydrogen bonds in protein–ligand binding
Pierce, A. C., Sandretto, K. L., & Bemis, G. W. (2002) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.10259 · Cited by 226 · see citing works
Morelli, X., Bourgeas, R., & Roche, P. (2011) · Current Opinion in Chemical Biology · doi:10.1016/j.cbpa.2011.05.024 · Cited by 223 · see citing works
Fast, accurate semiempirical molecular orbital calculations for macromolecules
Dixon, S. L., & Merz, K. M. (1997) · The Journal of Chemical Physics · doi:10.1063/1.474386 · Cited by 217 · see citing works
Inclusion of Solvation in Ligand Binding Free Energy Calculations Using the Generalized-Born Model
Zou, X., Yaxiong, & Kuntz, I. D. (1999) · Journal of the American Chemical Society · doi:10.1021/ja984102p · Cited by 214 · see citing works
Raha, K., & Merz, K. M. (2005) · Journal of Medicinal Chemistry · doi:10.1021/jm048973n · Cited by 209 · see citing works
D3R grand challenge 2015: evaluation of protein-ligand pose and affinity predictions.
Gathiaka, S., Liu, S., Chiu, M., Yang, H., Stuckey, J. A., Kang, Y. N., Delproposto, J., Kubish, G., Dunbar, J. B., Carlson, H. A., et al. (2016) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-016-9946-8 · Cited by 207 · see citing works
Scholz, C., Knorr, S., Hamacher, K., & Schmidt, B. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci500681r · Cited by 206 · see citing works
Cao, Y., & Li, L. (2014) · Bioinform. · doi:10.1093/bioinformatics/btu104 · Cited by 201 · see citing works
Bauer, M. R., Ibrahim, T. M., Vogel, S. M., & Boeckler, F. M. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci400115b · Cited by 200 · see citing works
Gaieb, Z., Liu, S., Gathiaka, S., Chiu, M., Yang, H., Shao, C., Feher, V. A., Walters, W. P., Kuhn, B., Rudolph, M. G., et al. (2018) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-017-0088-4 · Cited by 193 · see citing works
A Comprehensive Docking and MM/GBSA Rescoring Study of Ligand Recognition upon Binding Antithrombin
Zhang, X., Pérez‐Sánchez, H., & Lightstone, F. C. (2017) · Current Topics in Medicinal Chemistry · doi:10.2174/1568026616666161117112604 · Cited by 192 · see citing works
Poornima, C. S., & Dean, P. M. (1995) · Journal of Computer-Aided Molecular Design · doi:10.1007/bf00124321 · Cited by 191 · see citing works
A dynamic niching genetic algorithm strategy for docking highly flexible ligands
de Magalhães, C. S., Almeida, D. M., Barbosa, H. J. C., & Dardenne, L. E. (2014) · Information Sciences · doi:10.1016/j.ins.2014.08.002 · Cited by 190 · see citing works
Self-citationLi, G., Yang, L., Wang, W., Li, L., & Yang, S. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci300493w · Cited by 181 · see citing works
Ligand binding to proteins: The binding landscape model
Miller, D. W., & Dill, K. A. (1997) · Protein Science · doi:10.1002/pro.5560061011 · Cited by 178 · see citing works
Zilian, D., & Sotriffer, C. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci400120b · Cited by 174 · see citing works
Potential and Limitations of Ensemble Docking
Korb, O., Olsson, T. S. G., Bowden, S. J., Hall, R. J., Verdonk, M. L., Liebeschuetz, J. W., & Cole, J. C. (2012) · Journal of Chemical Information and Modeling · doi:10.1021/ci2005934 · Cited by 169 · see citing works
Rognan, D., Lauemøller, S. L., Holm, A., Buus, S., & Tschinke, V. (1999) · Journal of Medicinal Chemistry · doi:10.1021/jm9910775 · Cited by 165 · see citing works
The Taxonomy of Covalent Inhibitors
Tuley, A., & Fast, W. L. (2018) · Biochemistry · doi:10.1021/acs.biochem.8b00315 · Cited by 163 · see citing works
Glossary of terms used in computational drug design, part II (IUPAC Recommendations 2015)
Martin, Y. C., Abagyan, R. A., Ferenczy, G. G., Gillet, V. J., Oprea, T. I., Ulander, J., Winkler, D. A., & Zefirov, N. S. (2016) · Pure and Applied Chemistry · doi:10.1515/pac-2012-1204 · Cited by 161 · see citing works
Theory and Applications of Covalent Docking in Drug Discovery: Merits and Pitfalls
Kumalo, H., Bhakat, S., & Soliman, M. (2015) · Molecules · doi:10.3390/molecules20021984 · Cited by 158 · see citing works
Baum, B., Muley, L., Smolinski, M., Heine, A., Hangauer, D., & Klebe, G. (2010) · Journal of Molecular Biology · doi:10.1016/j.jmb.2010.02.007 · Cited by 157 · see citing works
Nguyen, D. D., Cang, Z., Wu, K., Wang, M., Cao, Y., & Wei, G.-W. (2019) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-018-0146-6 · Cited by 155 · see citing works
Fischer, M., Coleman, R. G., Fraser, J. S., & Shoichet, B. K. (2014) · Nature Chemistry · doi:10.1038/nchem.1954 · Cited by 149 · see citing works
Koebel, M. R., Schmadeke, G., Posner, R. G., & Sirimulla, S. (2016) · Journal of Cheminformatics · doi:10.1186/s13321-016-0139-1 · Cited by 148 · see citing works
Beware of machine learning-based scoring functions-on the danger of developing black boxes.
Gabel, J., Desaphy, J., & Rognan, D. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci500406k · Cited by 143 · see citing works
Computational tools to model halogen bonds in medicinal chemistry.
Ford, M. C., & Ho, P. S. (2016) · Journal of Medicinal Chemistry · doi:10.1021/acs.jmedchem.5b00997 · Cited by 143 · see citing works
Structure-Based Virtual Ligand Screening: Recent Success Stories
Villoutreix, B. O., Eudes, R., & Miteva, M. A. (2009) · Combinatorial Chemistry & High Throughput Screening · doi:10.2174/138620709789824682 · Cited by 139 · see citing works
Ouyang, X., Zhou, S., Su, C. T. T., Ge, Z., Li, R., & Kwoh, C. K. (2013) · Journal of Computational Chemistry · doi:10.1002/jcc.23136 · Cited by 134 · see citing works
Damm-Ganamet, K. L., Smith, R. D., Dunbar, J. B., Stuckey, J. A., & Carlson, H. A. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci400025f · Cited by 133 · see citing works
Docking Screens: Right for the Right Reasons?
Kolb, P., & Irwin, J. J. (2009) · Current Topics in Medicinal Chemistry · doi:10.2174/156802609789207091 · Cited by 132 · see citing works
Understanding the challenges of protein flexibility in drug design.
Antunes, D. A., Devaurs, D., & Kavraki, L. E. (2015) · Expert Opinion on Drug Discovery · doi:10.1517/17460441.2015.1094458 · Cited by 131 · see citing works
Kurkcuoglu, Z., Koukos, P. I., Citro, N., Trellet, M. E., Rodrigues, J., Moreira, I. S., Roel‐Touris, J., Melquiond, A. S. J., Geng, C., Schaarschmidt, J., et al. (2017) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-017-0049-y · Cited by 130 · see citing works
Assessing protein–ligand interaction scoring functions with the CASF-2013 benchmark
Li, Y., Su, M., Liu, Z., Li, J., Liu, J., Han, L., & Wang, R. (2018) · Nature Protocols · doi:10.1038/nprot.2017.114 · Cited by 125 · see citing works
Spyrakis, F., & Cavasotto, C. N. (2015) · Archives of Biochemistry and Biophysics · doi:10.1016/j.abb.2015.08.002 · Cited by 122 · see citing works
SFCscore: Scoring functions for affinity prediction of protein–ligand complexes
Sotriffer, C., Sanschagrin, P. C., Matter, H., & Klebe, G. (2008) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.22058 · Cited by 118 · see citing works
Chaput, L., Martinez‐Sanz, J., Saettel, N., & Mouawad, L. (2016) · Journal of Cheminformatics · doi:10.1186/s13321-016-0167-x · Cited by 118 · see citing works
CSM-lig: a web server for assessing and comparing protein-small molecule affinities
Pires, D. E. V., & Ascher, D. B. (2016) · Nucleic Acids Research · doi:10.1093/nar/gkw390 · Cited by 115 · see citing works
Krämer, C., & Gedeck, P. (2010) · Journal of Chemical Information and Modeling · doi:10.1021/ci100264e · Cited by 114 · see citing works
Ban, F., Dalal, K., Li, H., LeBlanc, E., Rennie, P. S., & Cherkasov, A. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00137 · Cited by 109 · see citing works
Decoys Selection in Benchmarking Datasets: Overview and Perspectives
Réau, M. F., Langenfeld, F., Zagury, J., Lagarde, N., & Montès, M. (2018) · Frontiers in Pharmacology · doi:10.3389/fphar.2018.00011 · Cited by 108 · see citing works
Rationalizing Tight Ligand Binding through Cooperative Interaction Networks
Kuhn, B., Fuchs, J. E., Reutlinger, M., Ståhl, M., & Taylor, N. R. (2011) · Journal of Chemical Information and Modeling · doi:10.1021/ci200319e · Cited by 107 · see citing works
Ericksen, S. S., Wu, H., Zhang, H., Michael, L. A., Newton, M. A., Hoffmann, F. M., & Wildman, S. A. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00153 · Cited by 107 · see citing works
Tufféry, P., & Derreumaux, P. (2011) · Journal of The Royal Society Interface · doi:10.1098/rsif.2011.0584 · Cited by 106 · see citing works
The role of protonation states in ligand-receptor recognition and binding.
Petukh, M., Stefl, S., & Alexov, E. (2013) · Current Pharmaceutical Design · doi:10.2174/1381612811319230004 · Cited by 101 · see citing works
Recent Updates on Computer-aided Drug Discovery: Time for a Paradigm Shift
Usha, T., Shanmugarajan, D., Goyal, A. K., et al. (2018) · Current Topics in Medicinal Chemistry · doi:10.2174/1568026618666180101163651 · Cited by 100 · see citing works
The impact of in silico screening in the discovery of novel and safer drug candidates
Rognan, D. (2017) · Pharmacology & Therapeutics · doi:10.1016/j.pharmthera.2017.02.034 · Cited by 99 · see citing works
Terp, G. E., Johansen, B. N., Christensen, I. T., & Jørgensen, F. S. (2001) · Journal of Medicinal Chemistry · doi:10.1021/jm001090l · Cited by 97 · see citing works
Can we use docking and scoring for hit-to-lead optimization?
Enyedy, I. J., & Egan, W. J. (2008) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-007-9165-4 · Cited by 97 · see citing works
Systematic exploitation of multiple receptor conformations for virtual ligand screening.
Bottegoni, G., Rocchia, W., Rueda, M., Abagyan, R., & Cavalli, A. (2011) · PLoS ONE · doi:10.1371/journal.pone.0018845 · Cited by 96 · see citing works
Vogel, S., Bauer, M. R., & Boeckler, F. M. (2011) · Journal of Chemical Information and Modeling · doi:10.1021/ci2001549 · Cited by 94 · see citing works
Blaber, M., Lindstrom, J. D., Gassner, N., Xu, J., Heinz, D. W., & Matthews, B. W. (1993) · Biochemistry · doi:10.1021/bi00093a013 · Cited by 93 · see citing works
Low-Quality Structural and Interaction Data Improves Binding Affinity Prediction via Random Forest
Li, H., Leung, K., Wong, M., & Ballester, P. J. (2015) · Molecules · doi:10.3390/molecules200610947 · Cited by 92 · see citing works
dMM-PBSA: A New HADDOCK Scoring Function for Protein-Peptide Docking
Spiliotopoulos, D., Kastritis, P. L., Melquiond, A. S. J., Bonvin, A. M. J. J., Musco, G., Rocchia, W., & Spitaleri, A. (2016) · Frontiers in Molecular Biosciences · doi:10.3389/fmolb.2016.00046 · Cited by 91 · see citing works
Li, H., Leung, K., Wong, M., & Ballester, P. J. (2014) · BMC Bioinformatics · doi:10.1186/1471-2105-15-291 · Cited by 90 · see citing works
Modeling covalent-modifier drugs.
Awoonor-Williams, E., Walsh, A. G., & Rowley, C. N. (2017) · Biochimica et Biophysica Acta (BBA) - Proteins and Proteomics · doi:10.1016/j.bbapap.2017.05.009 · Cited by 90 · see citing works
WScore: a flexible and accurate treatment of explicit water molecules in ligand-receptor docking.
Murphy, R. B., Repasky, M. P., Greenwood, J. R., Tubert-Brohman, I., Jerome, S., Annabhimoju, R., Boyles, N. A., Schmitz, C. D., Abel, R., Farid, R., et al. (2016) · Journal of Medicinal Chemistry · doi:10.1021/acs.jmedchem.6b00131 · Cited by 89 · see citing works
Divergent modes of enzyme inhibition in a homologous structure-activity series.
Ferreira, R. S., Bryant, C., Ang, K. K. H., McKerrow, J. H., Shoichet, B. K., & Renslo, A. R. (2009) · Journal of Medicinal Chemistry · doi:10.1021/jm9009229 · Cited by 88 · see citing works
Ashtawy, H. M., & Mahapatra, N. R. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00309 · Cited by 88 · see citing works
Supervised Consensus Scoring for Docking and Virtual Screening
Teramoto, R., & Fukunishi, H. (2007) · Journal of Chemical Information and Modeling · doi:10.1021/ci6004993 · Cited by 86 · see citing works
Lagarde, N., Zagury, J.-F., & Montes, M. (2015) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.5b00090 · Cited by 85 · see citing works
Why and how have drug discovery strategies in pharma changed? What are the new mindsets?
Mignani, S. M., Huber, S., Tomás, H., Rodrigues, J., & Majoral, J. (2015) · Drug Discovery Today · doi:10.1016/j.drudis.2015.09.007 · Cited by 83 · see citing works
Jain, T., & Jayaram, B. (2005) · FEBS Letters · doi:10.1016/j.febslet.2005.10.031 · Cited by 82 · see citing works
Ashtawy, H. M., & Mahapatra, N. R. (2014) · IEEE Transactions on Computational Biology and Bioinformatics · doi:10.1109/tcbb.2014.2351824 · Cited by 81 · see citing works
Water, water, everywhere… It’s time to stop and think.
Bodnarchuk, M. S. (2016) · Drug Discovery Today · doi:10.1016/j.drudis.2016.05.009 · Cited by 81 · see citing works
Quantum Chemical Approaches in Structure-Based Virtual Screening and Lead Optimization
Cavasotto, C. N., Adler, N. S., & Aucar, M. G. (2018) · Frontiers in Chemistry · doi:10.3389/fchem.2018.00188 · Cited by 81 · see citing works
Abagyan, R. A., Batalov, S., Cardozo, T. J., Totrov, M. M., Webber, J., & Zhou, Y. (1997) · Proteins Structure Function and Bioinformatics · doi:10.1002/(sici)1097-0134(1997)1+ · Cited by 78 · see citing works
Targeted scoring functions for virtual screening.
Seifert, M. H. J. (2009) · Drug Discovery Today · doi:10.1016/j.drudis.2009.03.013 · Cited by 76 · see citing works
Rapid measurement of inhibitor binding kinetics by isothermal titration calorimetry
Trani, J. M. D., Cesco, S. D., O’Leary, R., Plescia, J., do Nascimento, C. J., Moitessier, N., & Mittermaier, A. (2018) · Nature Communications · doi:10.1038/s41467-018-03263-3 · Cited by 74 · see citing works
Protein flexibility in drug discovery: from theory to computation.
Buonfiglio, R., Recanatini, M., & Masetti, M. (2015) · ChemMedChem · doi:10.1002/cmdc.201500086 · Cited by 73 · see citing works
A new class of models for computing receptor-ligand binding affinities
Gilson, M. K., Given, J. A., & Head, M. S. (1997) · Chemistry & Biology · doi:10.1016/s1074-5521(97)90251-9 · Cited by 72 · see citing works
Chaput, L., Martinez‐Sanz, J., Quiniou, É., Rigolet, P., Saettel, N., & Mouawad, L. (2016) · Journal of Cheminformatics · doi:10.1186/s13321-016-0112-z · Cited by 71 · see citing works
Wang, Y., Guo, Y., Kuang, Q., Pu, X., Ji, Y., Zhang, Z., & Li, M. (2015) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-014-9827-y · Cited by 69 · see citing works
Chaput, L., & Mouawad, L. (2017) · Journal of Cheminformatics · doi:10.1186/s13321-017-0227-x · Cited by 67 · see citing works
The effect of ligand-based tautomer and protomer prediction on structure-based virtual screening.
Kalliokoski, T., Salo, H. S., Lahtela-Kakkonen, M., & Poso, A. (2009) · Journal of Chemical Information and Modeling · doi:10.1021/ci900364w · Cited by 66 · see citing works
Yang, Y., Lightstone, F. C., & Wong, S. E. (2013) · Expert Opinion on Drug Discovery · doi:10.1517/17460441.2013.749853 · Cited by 65 · see citing works
Ashtawy, H. M., & Mahapatra, N. R. (2015) · BMC Bioinformatics · doi:10.1186/1471-2105-16-s4-s8 · Cited by 62 · see citing works
Entropy calculations on the molten globule state of a protein: Side‐chain entropies of α‐lactalbumin
Schäfer, H., Smith, L. J., Mark, A. E., et al. (2001) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.1166 · Cited by 61 · see citing works
GalaxyDock BP2 score: a hybrid scoring function for accurate protein–ligand docking
Baek, M., Shin, W., Chung, H. W., & Seok, C. (2017) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-017-0030-9 · Cited by 61 · see citing works
Practices in Molecular Docking and Structure-Based Virtual Screening
dos Santos, R. N., Ferreira, L. L. G., & Andricopulo, A. D. (2018) · Methods in molecular biology · doi:10.1007/978-1-4939-7756-7_3 · Cited by 61 · see citing works
Comparative assessment of machine-learning scoring functions on PDBbind 2013
Khamis, M. A., & Gomaa, W. E. (2015) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2015.06.021 · Cited by 59 · see citing works
Danishuddin, M., & Khan, A. U. (2014) · Methods · doi:10.1016/j.ymeth.2014.10.019 · Cited by 58 · see citing works
Park, S., Kufareva, I., & Abagyan, R. A. (2010) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-010-9362-4 · Cited by 58 · see citing works
Li, H., Peng, J., Leung, Y., Leung, K., Wong, M., Lü, G., & Ballester, P. J. (2018) · Biomolecules · doi:10.3390/biom8010012 · Cited by 57 · see citing works
Shedding light on important waters for drug design: simulations versus grid-based methods.
Bucher, D., Stouten, P., & Triballeau, N. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00642 · Cited by 57 · see citing works
Ligand-biased ensemble receptor docking (LigBEnD): a hybrid ligand/receptor structure-based approach
Lam, P. C. ‐., Abagyan, R. A., & Totrov, M. M. (2017) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-017-0058-x · Cited by 56 · see citing works
Robust scoring functions for protein-ligand interactions with quantum chemical charge models.
Wang, J.-C., Lin, J.-H., Chen, C.-M., Perryman, A. L., & Olson, A. J. (2011) · Journal of Chemical Information and Modeling · doi:10.1021/ci200220v · Cited by 54 · see citing works
Docking and High Throughput Docking: Successes and the Challenge of Protein Flexibility
Cavasotto, C. N., & Singh, N. P. (2008) · Current Computer - Aided Drug Design · doi:10.2174/157340908785747474 · Cited by 54 · see citing works
Smith, R. D., Damm-Ganamet, K., Dunbar, J., Ahmed, A., Chinnaswamy, K., Delproposto, J., Kubish, G., Tinberg, C., Khare, S., Dou, J., et al. (2015) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.5b00387 · Cited by 54 · see citing works
Kar, P., Lipowsky, R., & Knecht, V. (2013) · The Journal of Physical Chemistry B · doi:10.1021/jp3085292 · Cited by 53 · see citing works
Applications and Success Stories in Virtual Screening
Matter, H., & Sotriffer, C. (2011) · Methods and principles in medicinal chemistry · doi:10.1002/9783527633326.ch12 · Cited by 52 · see citing works
Ashtawy, H. M., & Mahapatra, N. R. (2012) · IEEE Transactions on Computational Biology and Bioinformatics · doi:10.1109/tcbb.2012.36 · Cited by 52 · see citing works
GFscore: a general nonlinear consensus scoring function for high-throughput docking.
Betzi, S., Suhre, K., Chétrit, B., Guerlesquin, F., & Morelli, X. (2006) · Journal of Chemical Information and Modeling · doi:10.1021/ci0600758 · Cited by 51 · see citing works
Liu, S., Fu, R., Zhou, L., & Chen, S. (2012) · PLoS ONE · doi:10.1371/journal.pone.0038086 · Cited by 50 · see citing works
Derivation of sensitive discrimination potential for virtual ligand screening
Totrov, M. M., & Abagyan, R. A. (1999) · doi:10.1145/299432.299509 · Cited by 50 · see citing works
Cramer, J., Krimmer, S. G., Heine, A., & Klebe, G. (2017) · Journal of Medicinal Chemistry · doi:10.1021/acs.jmedchem.7b00490 · Cited by 50 · see citing works
Ballester, P. J., & Mitchell, J. B. O. (2011) · Journal of Chemical Information and Modeling · doi:10.1021/ci200057e · Cited by 48 · see citing works
Implementing QM in docking calculations: is it a waste of computational time?
Adeniyi, A. A., & Soliman, M. E. S. (2017) · Drug Discovery Today · doi:10.1016/j.drudis.2017.06.012 · Cited by 48 · see citing works
Customized versus universal scoring functions
Logean, A., Sette, A., & Rognan, D. (2001) · Bioorganic & Medicinal Chemistry Letters · doi:10.1016/s0960-894x(01)00021-x · Cited by 46 · see citing works
On-the-fly QM/MM docking with attracting cavities.
Chaskar, P., Zoete, V., & Röhrig, U. F. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.6b00406 · Cited by 46 · see citing works
Ding, B., Wang, J., Li, N., & Wang, W. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci300508m · Cited by 45 · see citing works
Flagged first, then by citations
Research works printed but not linked: 3
- Abu-Mostafa, 2012, Learning From Data.
- Friesner, Extra precision glide: docking and scoring incorporating a model of hydrophobic enclosure for protein-ligand complexes., J. Med. Chem., 49, 6177, 10.1021/jm051256o · DOI
- Guedes, 2016, Development of empirical scoring functions for predicting protein-ligand binding affinity., Soc. Bras. Bioquim. Biol. Mol., 1
Cited by
Sorted by citations
Computational and artificial intelligence-based methods for antibody development
Kim, J., McFee, M., Fang, Q., Abdin, O., & Kim, P. M. (2023) · Trends in Pharmacological Sciences · doi:10.1016/j.tips.2022.12.005 · Cited by 178 · see citing works
- Authors
- Guedes, I. A., Pereira, F. S. S., & Dardenne, L. E.
- Affiliations
- Laboratório Nacional de Computação Científica (BR)
- Corresponding author
- Dardenne, Laurent E.
- Venue
- Frontiers in Pharmacology
- Publisher
- Frontiers Media SA
- Published
- 2018-09-24
- Volume
- 9
- Pages
- 1089-1089
- ISSN
- 1663-9812 (ISSN-L)
- PMC
- PMC6165880
- OpenAlex
- W2893435209
- Cited by
- 331
- References
- 239
- Citations by year
- 2026: 38 · 2025: 45 · 2024: 41 · 2023: 52 · 2022: 50 · 2021: 47 · 2020: 36 · 2019: 19 · 2018: 1 · 2012: 1
- FWCI
- 2.07
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- Frontiers in Pharmacology (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- Funders
- Conselho Nacional de Desenvolvimento Científico e Tecnológico
- Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro
- Topics
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › vaccines and immunoinformatics approaches
- Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Advanced Chemical Physics Studies
- Keywords
- empirical scoring function, structure-based virtual screening, molecular docking, binding affinity prediction, pose prediction, solvent effects, machine learning, quantum mechanics, lead compound design, compound ranking
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
