Distance-scaled, finite ideal-gas reference state improves structure-derived potentials of mean force for structure selection and stability prediction

“They include VERIFY3D (Luthy et al., 1992), PROSAII (Sippl, 1993), HARMONY (Topham et al., 1994), ANOLEA (Melo and Feytmans, 1998), and DFIRE (Zhou and Zhou, 2002).”

“template was used for the modeling. correct equivalences, that is, They include VERIFY3D (Luthy et al., 1992), Prosa2003 (Sippl, 1993; Wiederstein and Sippl, 2007), HARMONY (Topham et al., 1994), ANOLEA (Melo and Feytmans, 1998), DFIRE (Zhou and Zhou, 2002), DOPE (Shen and Sali, 2006), SOAP (Dong et al., 2013), QMEAN…”

“They include VERIFY3D (Luthy et al., 1992), PROSAII (Sippl, 1993), HARMONY (Topham et al., 1994), ANOLEA (Melo and Feytmans, 1998), and DFIRE (Zhou and Zhou, 2002).”

“The performance of QMEAN with respect to other stateof-the-art methods such as ProSA (Sippl, 1993) and DFIRE (Zhou and Zhou, 2002) has also been recently assessed in an independent study (Rykunov and Fiser, 2010).”

“Therefore, entries from the Repository can be directly submitted to the Workspace for quality assessment using different global and local quality scores such as DFire ( 45 ), ProQRes (46) or QMEAN (47).”

“Statistical potential approaches [16, 18, 19, 20, 21, 22, 23, 24, 25] derive potential functions using statistical analysis of the environmental propensities, substitution frequencies, and correlations of contacting residues in solved tertiary structures.”

“Edfire is the pairwise statistical potential from DFIRE (21), and ELJ is the pairwise physicsbased Lennard-Jones 12-6 potential (22).”

“…from data of known protein structures and are usually formalized as either distance-dependent or -independent pairwise potentials of mean force.12 –20 Alternatively, statistical potentials have been derived for other structural features such as torsion angles21–25 and solvent accessibility.25 –27…”

“…(Luthy et al., 1992), Prosa, 2003 (Sippl, 1993; Wiederstein and Sippl, 2007), HARMONY (Topham et al., 1994), ANOLEA (Melo and Feytmans, 1998), DFIRE (Zhou and Zhou, 2002), DOPE (Shen and Sali, 2006), SOAP (Dong et al., 2013), QMEAN local (Benkert et al., 2011), and TSVMod (Eramian et al., 2008).”

“We may distinguish the approaches based on statistical potentials extracted from datasets of protein structures [15-19], from those based on empirical potentials built from optimised combinations of various physical energy terms [20-23].”

“evaluated by using a normalized DFIRE score (63).”

“Dmutant exploits atom-based distance potentials with a finite ideal gas reference state (Zhou and Zhou, 2002).”

“The aim was to improve model assessment by statistical potentials of ProsaII (Sippl 1993) and DFIRE (Zhou and Zhou 2002), and by a composite model evaluation criterion (Melo et al.”

“Thus, during the last few years, several methods have been developed to predict the impact of nsSNPs on protein stability [10-16], protein function [17,18] and to detect their pathologic effect [19-29].”

“Large-scale interaction experiments are becoming more common, and general-purpose models to describe protein structures and energies are under development33, 35–37.”

“It is outside of the scope of this review to rank the performance of the existing tools (including Dmutant [73], MultiMutate [74], SCide [75], Scpred [76], SRide [77], nsSNPAnalyzer [78], Panther [79], PhD-SNP [6], SNAP [80] and SNPs&GO [81], among other well-known tools) because different benchmarking papers (see rev…”

“Some methods developed different energy functions, suited to compute the stability free energy [1-11], while other machine learning approaches [12-15].”

“After that, these different models are linked and steric clashes are removed by using the DFIRE potential functions (Yang and Zhou, 2008; Zhou and Zhou, 2002).”

“The first group of energy potentials is the knowledge-based atomic contact potentials by RW/RWplus (Zhang and Zhang, 2010) and Dfire/ dDfire (Yang and Zhou, 2008; Zhou and Zhou, 2002).”

“Structure based tools AUTO-MUTE [7], CUPSAT [8], Dmu-tant [9], FoldX [10], Eris [11], PoPMuSiC [12], SDM [13] or mCSM [14] usually perform better than the sequence based counterparts.”

“For example, in dDFIRE, Yang andZhou (31,34) introduced intoDFIRE the orientation dependence of polar atom interactions (treated as dipoles), which includes hydrogen-bonding interactions, and some improvement over DFIRE (14) in refolding the protein terminal regions with secondary structures was observed.”

“A simple ideal gas approximation may be used, and many alternatives have been proposed [81,87–89].”

“…structural information describes the structural the protein stability change upon mutation (Capriotti et al. , , 2005; Schymkowitz et al. , 2005; Zhou and Zhou, 2002). of the most important features for the prediction of the impact acids are often conserved in protein families and so changes…”

“, 2005], DMUTANT [Zhou and Zhou, 2002], and PoPMuSic [Gilis and Rooman, 2000] calculate mutational free energy changes of the protein based on its 3D structure.”

“Zhang et al. (2004) presented a simplified version of this all-atom based force field (Zhou and Zhou, 2002) by using the center of mass of side chains.”

“RAPDF [9], KBP [10], DFIRE [11], DOPE [12], OPUS-PSP [13,14], free-rotating chain-based potential [15], or the more composite TASSER/ITASSER [16,17,18] and ROSETTA [19] potentials], which is derived from the statistical regularities [20] of the solved protein structures in the PDB library [21].”

“Some of these researched the physical potential [5-7], some were based on statistical potentials [6,8-13] and some on empirical approaches that combined physical and statistical potentials to confer how the protein stability would change upon mutations [14-18]; still others were based on machine learning theories, by…”

“The overall reasonableness of the model geometry and conformation were verified by a number of parameters, including pseudo-energies calculated by QMEAN (−1204 kJ/mol) (25) and DFIRE (−3250 kJ/mol) (26).”

“For the marmoset and ferret DPP4, the RBD/ DPP4 complex model was generated by merging the RBD domain (PDB ID: 4KR0, Chain B) with the DPP4 model, which was then subjected to the binding energy calculation using an all-atom distance-dependent pairwise statistical potential, DFIRE [38].”

“) The DFIRE-based potential derived from a structure database of single-chain proteins.(21) The two numbers in each cell represent the rank of native structure and the Z score, respectively.”

“The last required information is the mean and standard deviation values used in equation (9).”

“RMSE and variance for different linear combinations of methods Methods Parameters Median bootstrapped RMSE Variance A B All Stabilizing Destabilizing A B Shared ENCoM FoldX − 1.12 0.38 1.24 ± 0.14 1.45 ± 0.39 1.40 ± 0.17 0.12 0.24 0.05 ENCoM Rosetta − 1.14 0.46 1.32 ± 0.16 1.70 ± 0.41 1.43 ± 0.21 0.14 0.15 0.04 FoldX…”

“In the top 5% of the entire 1101-variant database, DFIRE, FoldX, and Discovery Studio identified different stabilized variants; only 11 out of 89 total variants with DD G < 2 1.0 kcal/mol were found by all three methods.”

“Differently from our approach, all the methods available use the molecular protein structure at different level of representation for predicting the G values after the protein mutation (Gilis and Rooman, 1997; Funahashi et al., 2001; Guerois et al., 2002; Kwasigroch et al., 2002; Zhou and Zhou, 2002).”

“A set of models for given target sequence is clustered together and the clusters are evaluated by MODELLER scoring function [43] and/or DFIRE energy [44].”

“However, most scoring functions using knowledge-based methods (Benkert et al., 2011; Luthy et al., 1992; Ray et al., 2012; Sippl, 1993; Zhou and Zhou, 2002) have been trained on soluble proteins.”

“Known structure-based algorithms include Dmutant [3], FoldX [4], I-Mutant2.”

“Empirical methods to calculate mutation-induced stability changes of proteins have been investigated intensively in the past years [20, 21, 22, 23, 24, 25, 26, 27, 28, 29].”

“58 SPIN was trained on 1532 non-redundant proteins and reaches a sequence identity of 30.3% on a test set containing 500 proteins.”

“…taking into account evolutionary information [16-18], stereochemical plausibility of the models [19,20] and the environment compatibility of their residues [21] as well as energy-based methods which include physics-based energy functions [22,23] and knowledge-based statistical potentials [24-29].”

“QMEAN [80] and DFire [84] were used to estimate global model quality.”

“Examples of statistical potentials include ProsaII (Sippl 1993a, b), ANOLEA (Melo and Feytmans 1997, 1998), and DFIRE (Zhou and Zhou 2002; Zhang et al. 2004).”

“(ii) Severe form of phenylketonuria: G46S; human phenylalanine hydroxylase (hPAH); [176] DFIRE [55, 177, 178] PoPMuSiC-2.”

“[22–27].”

“These techniques are derived either from knowledge-based statistical potentials (Gilis and Rooman, 1996, 1997; Hoppe and Schomburg, 2005; Kwasigroch et al., 2002; Meyerguz et al., 2007; Ota et al., 1995; Parthiban et al., 2007; Topham et al., 1997; Wang et al., 1998; Zhou and Zhou, 2002), or from physical descriptions…”

“To our knowledge, protein dynamics and flexibility have not been explicitly modeled inpreviousDDGpredictionmethods.Ignoringproteinflexibility limits the application of current prediction methods to a small range of mutations (Guerois et al., 2002; Zhou and Zhou, 2002 ).”

“These algorithms are able to predict the effect of the mutation on protein stability [13,14,19,23-26], protein functional activity [9,21], and insurgence of human pathologies [7,8,10-12,1518,20-22].”

“The development of more sophisticated scoring methods, taking into account various aspects of structural and sequence information in order to be able to judge the quality of obtained models (41–45), is currently matter of intensive research.”

“For all three statistical potentials of mean force (KBP, Lu and Skolnick 2001; RAPDF, Samudrala and Moult 1998; DFIRE, Zhou and Zhou 2002), the potential based on the side-chain center of mass (or center of geometry) is the most accurate one.”

“Some of them require the knowledge of the protein structure: AUTO-MUTE (Masso and Vaisman, 2008), CUPSAT (Parthiban et al., 2006), Dmutant (Zhou and Zhou, 2002), FoldX (Guerois et al., 2002), Eris (Yin et al., 2007), PoPMuSiC (Dehouck et al., 2009), SDM (Topham et al., 1997; Worth et al., 2011), mCSM (Pires et al., 20…”

“2002; McGuffin and Jones 2003; Melo and Sali 2007), (2) discriminating between the native and near-native states (Sippl 1993; Melo and Feytmans 1997, 1998; Park et al. 1997; Fiser et al. 2000; Lazaridis and Karplus 2000; Zhou and Zhou 2002; Seok et al. 2003; Tsai et al. 2003; Shen and Sali 2006), (3) selecting the mos…”

“Although these potentials describe their reference state as an ideal gas, in reality, this state is closer to an ideal solution of noninteracting uniform-size particles [26 ,27].”

“Our large-scale experimental test also indicates EPAD is much better than those context-independent distance-based pairwise potentials, such as DOPE (Shen and Sali, 2009), DFIRE (Zhou and Zhou, 2009) and RW (Zhang and Zhang, 2010) in ranking protein alignments (or templatebased models) generated by both threading and…”

“Some of the first methods to predict the impact of nsSNVs were based solely on structure.(54,55) They”

“We tested the performance of LightDock (using default parameters; see Methods) with DFIRE (Zhou and Zhou, 2002) scoring function (LightDock-DFIRE), as well as that of LightDock with a faster implementation of the pyDock (Cheng et al., 2007) scoring function (Supplementary Methods 1.6) called…”

“…total of 30 bins, 158 residue-specific atom types with identical atom types merged (e.g., NH1 and NH2 in Arg), while the original DFIRE used a variable bin width (2 Å for the first 2 Å, 0.5 Å from 2 Å and 8 Å, and 1 Å from 8 to 15 Å) and 167 residue-specific atom types (Zhou and Zhou 2002).”

“10,11 Another popular refinement approach is using knowledge-based potentials such as DFIRE,20–22 dDFIRE,11,23 and RAPDF22,24–26 for sampling and/or scoring.”

“61 is best suited for finite proteins.(6) The potential was generated by using a nonhomologous ( 30% homology) structural database of 1011 soluble proteins with resolution 2 Å.”

“, 2015), DFIRE (Zhou and Zhou, 2002), dDFIRE(Yang and Zhou, 2008), and Rosetta (Humphris and Kortemme, 2008).”

“Several potentials have been tested on the group of decoy sets D(1)str (30,47); the results are summarized in Table 3.”

“Force field-based algorithms were also introduced decades ago and are still being actively researched [13–15].”

“First, if we assume that atoms are placed randomly within the protein shell, a recovery function will be similar to the DFIRE and DOPE reference states based on the ideal gas assumption (Shen and Sali, 2006; Zhou and Zhou, 2002).”

“Boltzmann inversion, whereas molecules are finite systems composed of interacting particles (31,32).”