CoChat
Sign Up
Cited by 190Protein Structure and DynamicsComputational Drug Discovery MethodsMachine Learning in Bioinformatics

AlphaFold, Artificial Intelligence (AI), and Allostery

Open access
Nussinov, Ruth, Zhang, Mingzhen, Liu, Yonglan, Jang, Hyunbum·2022·The Journal of Physical Chemistry B
DOI 10.1021/acs.jpcb.2c04346PMC PMC9442638
Open this paper in CoChat — free account

Abstract

AlphaFold has burst into our lives. A powerful algorithm that underscores the strength of biological sequence data and artificial intelligence (AI). AlphaFold has appended projects and research directions. The database it has been creating promises an untold number of applications with vast potential impacts that are still difficult to surmise. AI approaches can revolutionize personalized treatments and usher in better-informed clinical trials. They promise to make giant leaps toward reshaping and revamping drug discovery strategies, selecting and prioritizing combinations of drug targets. Here, we briefly overview AI in structural biology, including in molecular dynamics simulations and prediction of microbiota-human protein-protein interactions. We highlight the advancements accomplished by the deep-learning-powered AlphaFold in protein structure prediction and their powerful impact on the life sciences. At the same time, AlphaFold does not resolve the decades-long protein folding challenge, nor does it identify the folding pathways. The models that AlphaFold provides do not capture conformational mechanisms like frustration and allostery, which are rooted in ensembles, and controlled by their dynamic distributions. Allostery and signaling are properties of populations. AlphaFold also does not generate ensembles of intrinsically disordered proteins and regions, instead describing them by their low structural probabilities. Since AlphaFold generates single ranked structures, rather than conformational ensembles, it cannot elucidate the mechanisms of allosteric activating driver hotspot mutations nor of allosteric drug resistance. However, by capturing key features, deep learning techniques can use the single predicted conformation as the basis for generating a diverse ensemble.

AI snapshot

Written from the paper’s full text.

TL;DR

AlphaFold has transformed protein structure prediction by leveraging deep learning and large sequence/structure databases, with major implications for structural biology, drug discovery, and microbiome–host interaction studies. However, because it outputs single ranked structures rather than conformational ensembles, it cannot directly resolve allostery, frustration, folding pathways, or intrinsically disordered protein ensembles.

Key findings

  • AlphaFold predicts 3D protein structures at or near experimental resolution; it achieved scores near 90 even on difficult targets in the 2020 CASP, was trained on PDB chains and sequence databases, and “illuminates half of the dark human proteins.” The PDB is described as nearing 200,000 experimentally determined structures, and over 180,000 protein structures are available in the PDB database.
  • AlphaFold does not solve the protein folding problem or identify folding pathways; its models are rigid and do not capture conformational mechanisms such as frustration and allostery, which are rooted in ensembles and controlled by dynamic distributions. It also does not generate ensembles of intrinsically disordered proteins/regions, instead describing them by low structural probabilities.
  • Because AlphaFold generates single ranked structures rather than conformational ensembles, it cannot directly elucidate mechanisms of allosteric activating driver hotspot mutations or allosteric drug resistance. Its predicted structures can, however, be submitted to MD simulations to generate ensembles, and AI/deep-learning methods can accelerate and enhance MD simulations.
  • AI approaches are also discussed for human–microbiome protein–protein interaction prediction, drug repurposing, and selecting drug target combinations; allosteric drugs are described as having “anchor” and “driver” atoms, where driver atoms initiate the allosteric signal by shifting receptor populations.

Methods

  • This is a review/overview/perspective article, not a primary experimental study; it synthesizes AI in structural biology, molecular dynamics, human–microbiome PPI prediction, drug discovery, and allostery.
  • It uses illustrative examples, including SHP2 (X-ray structure PDB ID: 4DGP missing flexible-region residues predicted by AlphaFold), B-Raf kinase OFF-state (PDB ID: 3SKC) and ON-state (PDB ID: 6UAN) conformations compared with an AlphaFold structure, and Cdc42/GEF mimicry examples (PDB IDs: 2WMO, 1GZS, 3GCG).
  • Main analysis is conceptual and literature-based, with figures; no new dataset, sample, statistical analysis, or systematic review methodology is reported.

Limitations

  • The paper states that AlphaFold cannot directly address allosteric mechanisms, population shifts, folding pathways, or IDP/IDR ensembles; it generates single ranked structures, and current allosteric binding-site prediction approaches address stable interactions at the site but not initiation of effective allosteric signals.
  • The paper states that even though MD simulations of AlphaFold models could address dynamics, at the required scales the cost is prohibitive, and a paradigm-shifting machine learning method is needed to model protein dynamics.
  • Evident from the text: as a brief overview/perspective, it provides no primary validation, no quantitative synthesis, and no systematic assessment of the AI methods discussed.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1021/acs.jpcb.2c04346

Cites

207 of 207 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 96 cited sources checked · 0 supported · 0 flagged

  • Findingnone given

    The authors have developed an algorithm and server to predict human host-microbial protein-protein interactions (HMIs) based on their protein structures, which can be experimental or modeled.

    “We have developed an algorithm and server to predict these human host–microbial PPIs (HMIs) based on their protein structures, which can be experimental or modeled.”
    Quote in the paperSubstantial claim
  • Premisecitation

    AlphaFold combines deep learning innovations to predict three-dimensional protein structures at or near experimental scale resolution.

    “AlphaFold has combined numerous deep learning innovations to predict the three-dimensional (3D) structures of proteins at or near experimental scale resolution”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [1] source

      Not checked in the free run.

    • cites [2] source

      Not checked in the free run.

    • cites [3] source

      Not checked in the free run.

  • Interpretationreasoning

    AlphaFold models do not capture conformational mechanisms such as frustration and allostery, which are rooted in ensembles and controlled by dynamic distributions.

    “The models that AlphaFold provides do not capture conformational mechanisms like frustration and allostery, which are rooted in ensembles, and controlled by their dynamic distributions.”
    Quote in the paperNot a substantial claim
  • Interpretationreasoning

    Prediction of the structures of intrinsically disordered proteins and regions is a problem where AlphaFold falls short.

    “prediction of the structures of intrinsically disordered proteins (IDPs) and regions (IDRs) is another problem where AlphFold falls short.”
    Quote in the paperNot a substantial claim
  • Interpretationreasoning

    AlphaFold can indirectly assist in unraveling mechanisms of allosteric hotspot mutations and allosteric drug discovery by providing rigid structures that can be submitted to MD simulations to generate ensembles.

    “The rigid structures that AlphaFold predicts can be submitted to MD simulations that generate such ensembles”
    Quote in the paperSubstantial claim
  • May combine two claims. 36
  • Classified as probably not a substantial claim. 38
  • The quote only partly states this claim. 2

  • May combine two claims.

    Interpretationcitation

    AlphaFold does not resolve the protein folding problem and does not identify folding pathways.

    “AlphaFold does not resolve the decades-long protein folding challenge, nor does it identify the folding pathways.”
    • Also stated in: § Protein Folding · plain
      “Neither AlphaFold nor broadly other protein structure prediction algorithms consider folding pathways.”
    Quote in the paperSubstantial claim0 of 7 sources checked
    • cites [51] source

      Not checked in the free run.

    • cites [52] source

      Not checked in the free run.

    • cites [53] source

      Not checked in the free run.

    • cites [54] source

      Not checked in the free run.

    • cites [55] source

      Not checked in the free run.

    • cites [56] source

      Not checked in the free run.

    • cites [57] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    AlphaFold is trained on protein chains from the PDB and uses the input sequence to query protein sequence databases to construct a multiple sequence alignment.

    “It has been trained on protein chains from the PDB and uses the input sequence to query databases of protein sequences to construct a multiple sequence alignment.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    Deep learning techniques can use AlphaFold's single predicted conformation as the basis for generating a diverse ensemble.

    “However, by capturing key features, deep learning techniques can use the single predicted conformation as the basis for generating a diverse ensemble.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    AlphaFold's success has not led to a deeper mechanistic understanding of how a protein sequence folds, and it does not assist in folding a protein from its sequence.

    “However, its striking success has not led us to a deeper mechanistic understanding of exactly how a protein sequence folds, thus not assisting in the folding of a protein from its sequence.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Because AlphaFold generates single ranked structures rather than conformational ensembles, it cannot elucidate the mechanisms of allosteric activating driver hotspot mutations or allosteric drug resistance.

    “Since AlphaFold generates single ranked structures, rather than conformational ensembles, it cannot elucidate the mechanisms of allosteric activating driver hotspot mutations nor of allosteric drug resistance.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Allostery and signaling are properties of populations.

    “Allostery and signaling are properties of populations.”
    • Also stated in: § Introduction · plain
      “Allostery, where the signal propagates dynamically with the shifts in the populations, underlies regulation and thus cell life.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [118] source

      Not checked in the free run.

    • cites [119] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    AlphaFold, which predicts single ranked structures for a protein sequence, is unable to address directly allosteric mechanisms, which are based on populations of conformational states in ensembles.

    “AlphaFold, which predicts single ranked structures for a protein sequence, is unable to address directly allosteric mechanisms, which are based on the populations of conformational states in the ensembles.”
    Quote in the paperSubstantial claim0 of 10 sources checked
    • cites [101] source

      Not checked in the free run.

    • cites [102] source

      Not checked in the free run.

    • cites [103] source

      Not checked in the free run.

    • cites [104] source

      Not checked in the free run.

    • cites [105] source

      Not checked in the free run.

    • cites [106] source

      Not checked in the free run.

    • cites [107] source

      Not checked in the free run.

    • cites [108] source

      Not checked in the free run.

    • cites [109] source

      Not checked in the free run.

    • cites [110] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    AlphaFold does not generate ensembles of intrinsically disordered proteins and regions; it describes them by low structural probabilities.

    “AlphaFold also does not generate ensembles of intrinsically disordered proteins and regions, instead describing them by their low structural probabilities.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims. The quote only partly states this claim.

    Premisenone given

    Anchor atoms bind to the allosteric pocket without changing the conformation of the binding site, stabilizing ligand binding.

    “The anchor atoms bind to the allosteric pocket, without changing the conformation of the binding site.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Allosteric drugs can be constituted of anchor and driver atoms.

    “We suggested that allosteric drugs can constitute of “anchors” and “drivers” atoms”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Allosteric drugs have higher specificity and lower toxicity due to targeting nonconserved allosteric sites.

    “Due to its higher specificity and consequently lower toxicity, which results from targeting nonconserved allosteric sites, allostery also increasingly features in allosteric drugs.”
    Quote in the paperNot a substantial claim0 of 6 sources checked
    • cites [120] source

      Not checked in the free run.

    • cites [121] source

      Not checked in the free run.

    • cites [122] source

      Not checked in the free run.

    • cites [123] source

      Not checked in the free run.

    • cites [124] source

      Not checked in the free run.

    • cites [125] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Other AI-based strategies can assist directly in allosteric efforts, most effectively by accelerating and enhancing MD simulations.

    “other AI-based strategies can assist directly in such efforts, most effectively via accelerating and enhancing MD simulations.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Effective allosteric sites require both stable interactions at the site and initiation of effective allosteric signals.

    “the effectiveness of allosteric sites is determines by both stable interactions at the site, which is something that AI can help with, and initiation of effective allosteric signals, which would be more challenging.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    Driver atoms pull or push atoms in the protein pocket, initiating the allosteric signal that shifts the receptor population from the inactive to the active state.

    “The binding of driver atoms “pulls” or “pushes” atoms in the protein pocket. This initiates the allosteric signal, which shifts the receptor population from the inactive to the active state.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Orthosteric drugs block the active site; allosteric drugs alter the population of the active state through binding at a site far away.

    “Orthosteric drugs block the active site; allosteric drugs alter the population of the active state of the protein, including the active site, through binding at a site far away.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [126] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Disordered proteins and regions are characterized by broad and heterogeneous ensembles with small differences in relative conformational stabilities and low barriers.

    “Disordered proteins (regions) are characterized by broad and heterogeneous ensembles where the differences in the relative conformational stabilities are small, or even minor and the barriers are low.”
    Quote in the paperSubstantial claim0 of 9 sources checked
    • cites [127] source

      Not checked in the free run.

    • cites [128] source

      Not checked in the free run.

    • cites [129] source

      Not checked in the free run.

    • cites [130] source

      Not checked in the free run.

    • cites [131] source

      Not checked in the free run.

    • cites [132] source

      Not checked in the free run.

    • cites [133] source

      Not checked in the free run.

    • cites [134] source

      Not checked in the free run.

    • cites [135] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    AI strategies can handle population shifts but will need to go beyond prediction of stabilizing interactions.

    “AI strategies can but will need to go beyond prediction of stabilizing interactions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The folding paradigm stipulated that unfolded molecules spontaneously fold into the same shape, with the linear amino acid sequence specifying the folded native state.

    “The folding paradigm stipulated that unfolded molecules will always spontaneously fold into the same shape; that is, the linear amino acid sequence specifies a protein’s folded native state.”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [58] source

      Not checked in the free run.

    • cites [142] source

      Not checked in the free run.

    • cites [143] source

      Not checked in the free run.

    • cites [144] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Anfinsen's thermodynamic hypothesis states that the native structure of a protein is its thermodynamically most stable structure, determined only by its amino acid sequence and conditions, with kinetics playing no role.

    “The landmark thermodynamic hypothesis of Christian Anfinsen and his colleagues140,141 stated that the native structure of a protein is its thermodynamically most stable structure, and it is determined only by its amino acid sequence and the conditions it is at, with kinetics playing no role.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [140] source

      Not checked in the free run.

    • cites [141] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Unlike quantum computing, for AI and data-driven computing, the technological innovations at the requisite scales are already at hand.

    “Unlike quantum computing, in the case of AI and data-driven computing, the technological innovations at the requisite scales are already at hand.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The conformations of intrinsically disordered proteins and regions interconvert, leading to low probabilities of AlphaFold reliably capturing the most favored conformation or the conformational distribution.

    “The conformations interconvert, leading to low probabilities of AlphaFold’s reliably capturing those most favored, or the conformational distribution.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    AI-powered algorithms fed vast compiled data and enabled by massive compute power are propelling a revolution in computational biology.

    “AI-powered algorithms, which are fed vast compiled data, and enabled by the emerging massive compute power are propelling a revolution in computational biology”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Levinthal argued that there is no need to search the vast folding space because the energy landscape is funnel-like and can guide sampling toward the biological conformational basin.

    “Levinthal argued that there is no need to search this vast space since the energy landscape is funnel-like, rather than flat, and thus can guide sampling toward the biological conformational basin.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [51] source

      Not checked in the free run.

    • cites [153] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    For prediction algorithms sampling backbone states, the search space size grows exponentially with chain length, becoming an impossibility.

    “For prediction algorithms’ sampling backbone states, the search space size grows exponentially with chain length, becoming an impossibility.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Single sequence-based prediction considers hydrogen bonds, ion pairs, van der Waals attractions, and chiefly water-mediated hydrophobic interactions, with the hydrophobic effect driving protein folding.

    “Single sequence-based prediction in solution considers forces related to hydrogen bonds, ion pairs, van der Waals attractions,139 and chiefly water-mediated hydrophobic interactions, with the hydrophobic effect the driving force for protein folding.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [139] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Levinthal conceptualized the key problem facing protein folding and prediction algorithms as the vast time scales for a protein to search folding space and reach its most stable native state under biological conditions.

    “Already early on, Cyrus Levinthal conceptualized the key problem facing the protein and the prediction algorithms:152 the vast time scales for the protein to search the folding space and reach its most stable native state under biological conditions.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [152] Levinthal C., 1969

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    A static description cannot capture function, including protein activation, high-affinity activator binding shifts, allosteric rescue mutations, allosteric drugs blocking the active site, and overcoming countering mutations.

    “Yet, a static description cannot capture function.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisenone given

    AlphaFold's remarkable success, with scores near 90 even in difficult CASP 2020 targets, was influenced by training on all PDB structures and on structures it predicted, and it uses structure and correlation data to predict contacting amino acid pairs and all pairwise distances.

    “Its remarkable success (with scores of near 90 even in the difficult targets in the 2020 Critical Assessment of Protein Structure Prediction, CASP) was influenced by its training not only on all the PDB structures but also on structures it predicted, and it uses the structure and correlation data to predict the pairs of amino acids that are in contact as well as all amino acid pairwise distances.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    AlphaFold models can serve as input to generate ensembles, for example by MD simulations, which if carried out at sufficiently long time scales, in parallel, should be able to produce them.

    “However, the models it produces can serve as input to generate ensembles, for example by MD simulations, which, if carried out at sufficiently longtime scales, in parallel, it should be able to produce.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Frustration does not create new conformations; it alters the number of molecules populating a conformation.

    “Frustration does not create new conformations; instead, it alters the number of molecules populating it.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [173] Christensen N. G., 1989, Tidsskr Nor Laegeforen

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    AlphaFold is not the first machine-learning model for protein structure prediction.

    “AlphaFold is not the first in being a machine-learning model.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    AlphaFold falls short in addressing which structural states exist for a given protein and what the population of each state is.

    “Still, questions remain, such as which structural states exist for a give protein, and what is the population of each state. Addressing these questions is vital to relate protein structure to function. This is where AlphaFold falls short.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    To date, AlphaFold illuminates half of the dark human proteins.

    “To date, AlphaFold illuminates half of the dark human proteins.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    Simulations can sample relevant states, enumerate possible state combinations in multistate models, and determine the population sizes for the states.

    “Simulations can sample the relevant states, can enumerate possible state combinations (multistate models), and can determine the population sizes for the states.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    AlphaFold predictions are unable to address allostery directly; only through sampling can this functional aim be accomplished.

    “Despite its transformative power and vast broad impact, the AlphaFold predictions are unable to address it directly. It is only through their sampling that this functional aim can be accomplished.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Biomolecules are dynamical objects that are always interconverting between structures with varying energies, and this is the origin of allosteric mechanisms.

    “biomolecules are dynamical objects that are always interconverting between a variety of structures with varying energies,166 and that this is the origin of allosteric mechanisms.”
    Quote in the paperSubstantial claim0 of 4 sources checked
    • cites [166] source

      Not checked in the free run.

    • cites [167] source

      Not checked in the free run.

    • cites [168] source

      Not checked in the free run.

    • cites [169] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    AlphaFold ensured that distances between amino acids satisfy the triangle inequality, saving time at intermediate steps.

    “It also ensured that the distances between the amino acids satisfy the triangle inequality, saving time at intermediate steps.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [164] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    To understand biological regulation, structure should be linked to function through protein ensembles in terms of populations and relative energies, which is the foundation of allostery.

    “To understand biological regulation, structure should be linked to function through protein ensembles in terms of populations and relative energies, which is the foundation of allostery”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Interpretationnone given

    AlphaFold predictions, like homology modeling, are rigid; flexibility is implicitly captured by absence or low confidence levels for certain regions, as in intrinsically disordered proteins.

    “The predictions that it yields, like those obtained by homology modeling, are rigid.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Deep learning methods can help MD simulations excel in efficiency and scales, with AI bridging deep learning technologies and simulations.

    “Deep learning methods can help MD simulations excel in their efficiency and scales, with AI bridging between deep learning technologies and simulations.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A deep learning framework with mixed classical and machine learning potentials (TorchMD) has been developed for molecular simulations.

    “In another example, a deep learning framework with mixed classical and machine learning potentials (TorchMD) has been developed for molecular simulations.188”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [188] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Deep learning involving variational autoencoders has been shown useful; the learned latent space has been employed to generate unsampled protein conformations, and simulations starting from these conformations accelerated sampling.

    “The learned latent space in the variational autoencoders has been employed to generate unsampled protein conformations, and simulations starting from these conformations accelerated the sampling.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [193] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    MD simulations need to perform extensive sampling of the conformational space that requires long time scales.

    “MD simulations need to perform extensive sampling of the conformational space that require long time scales.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The conformational selection mechanism theorizes that the energy surface hosts a very large number of conformations, the best-fitting one is selected, with subsequent minor induced fit optimization largely by side chains.

    “In contrast, the conformational selection mechanism167−169,185 theorizes that the energy surface hosts a very large number of conformations, and the one that fits best is selected, with subsequent minor induced fit optimization, largely by side chains.”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [167] source

      Not checked in the free run.

    • cites [168] source

      Not checked in the free run.

    • cites [169] source

      Not checked in the free run.

    • cites [185] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Machine learning has already been making a significant impact on the development of approximate methods for complex atomic systems.

    “machine learning has already been making a significant impact on the development of approximate methods for complex atomic systems.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [186] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    In the induced fit mechanism, ligands bind to the single open protein structure and interaction with a rigid binding partner induces a conformational change.

    “In an induced fit scenario, the ligands bind to the single “open” protein structure and the interaction between a protein and a rigid binding partner induces a conformational change in the protein.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [184] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    While there is a single conformation that the active enzyme should adopt for productive catalysis, there are multiple ways to inactivate it and thus many inactive states.

    “While there is a single conformation that the active enzyme should adopt for productive catalysis, there are multiple ways to inactivate it and thus many inactive states.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    AI and deep learning are being developed and applied to experimental determination and prediction of macromolecular structures and to protein-protein interactions.

    “AI and deep learning are also being developed and applied to experimental determination and prediction of macromolecular structures,198 as well as to PPIs.199,200”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [198] source

      Not checked in the free run.

    • cites [199] source

      Not checked in the free run.

    • cites [200] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Machine learning permits large-scale efficient and generalizable application and can address the complex dynamics of host-microbe relationships.

    “Machine learning permits both the large-scale efficient and generalizable application and addressing the complex dynamics of such relationships that the machine learning algorithms can decipher.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    If the structures of the human or microbe are unavailable, AlphaFold can generate them.

    “If the structures of the human or microbe are unable, AlphaFold can generate them.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Through MD simulations of models that AlphaFold produces, allostery can be accomplished as well.

    “Nevertheless, through MD simulations of models that AlphaFold produces, this aim can be accomplished as well.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Allosteric mechanisms fall into the category of what AlphaFold cannot accomplish.

    “Allosteric mechanisms fall into the latter category.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    In large-scale applications, AlphaFold can now be used to predict human host-microbial protein-protein interactions.

    “In large scale applications, AlphaFold can now be used toward this aim.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Evaluation metrics for PPI prediction are unclear and PPI networks are sparse.

    “Evaluation metrics are also unclear, the PPI networks are sparse, and more.66”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [66] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Premisenone given

    DeepMind's AlphaFold2 success in sequence-based protein structure prediction, the RoseTTAFold open-source counterpart, and the publicly available AlphaFold2 prediction of all human proteins are major steps that benefit the scientific community.

    “DeepMind’s AlphaFold2 success in sequence-based protein structure prediction,4 as well as the RoseTTAFold70 open-source counterpart, and the publicly available AlphaFold2 prediction of all human proteins203 are major steps that benefit the scientific community.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [4] source

      Not checked in the free run.

    • cites [70] source

      Not checked in the free run.

    • cites [203] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    A paradigm-shifting machine learning method is needed to model protein dynamics.

    “A paradigm-shifting machine learning method is needed to model protein dynamics.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Even though simulations would address the dynamics problem, at such scales, the cost is prohibitive.

    “Still, however, even though simulations would address this dynamics problem, at such scales, the cost is prohibitive.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationcitation

    AlphaFold and its underlying deep learning innovations have opened up the next frontiers in protein science, including precision medicine.

    “AlphaFold and its underlying deep learning innovations have opened up the next frontiers in protein science,164 including precision medicine.206”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [164] source

      Not checked in the free run.

    • cites [206] source

      Not checked in the free run.

58 of 58

  • The authors state this as an assumption.

    The crystal structure represents the minimum energy state, so the goodness of a predicted structure can be assessed by comparison with it.

    “Assuming that the crystal structure represents the minimum energy state, the “goodness” of the predicted structure can then be assessed by comparison with it.”

  • The paper relies on this without support.

    Human–microbiome protein–protein interactions play important roles in human health and disease.

    “These interactions play important roles in human health and disease.”

  • The paper relies on this without support.

    Microbes, bacteria, and viruses impact human health.

    “There is a rapid increase of data that microbes, bacteria, and viruses impact human health.”

  • The paper relies on this without support.

    Microbes can modulate human signaling and immune response by interacting with human proteins.

    “They can modulate human signaling and immune response by interacting with the human proteins.”

  • The paper relies on this without support.

    The interactions help pathogens elude and bypass immune defense, with pathogens hijacking host signaling.

    “The interactions help the pathogens to elude and bypass the immune defense, with the pathogens hijacking host signaling.”

  • The paper relies on this without support.

    Data including structural details are scarce and large-scale experimental detection is challenging.

    “Data, including structural details, are scarce and large-scale experimental detection is challenging.”

  • The paper relies on this without support.

    Data relating to host–microbe PPIs with 3D structures are sparse, causing problems in training and evaluating computational methods.

    “Data relating to host-microbe PPIs with 3D structures are sparse, thus facing a problem in training and evaluating the computational methods.”

  • The paper relies on this without support.

    There is a lack of gold standard test data set for PPI prediction.

    “Additional problems involve lack of gold standard test data set.”

  • The paper relies on this without support.

    To date, over 180,000 protein structures are available in the PDB database, open to all researchers worldwide.

    “To date, over 180 000 protein structures are available, open to all researchers across the world in the PDB database.”

  • The paper relies on this without support.

    Structures of pathogens are not within the PDB, and neither are many other structures essential to human health.

    “Still, structures of pathogens are not within them, and neither are many others, which are essential to human health.”
10 of 10

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?

  • AlphaFold successfully solved the long-standing problem of protein folding, which had hindered progress in biology and medicine for decades.

    “For instance, AlphaFold, an AI-powered algorithm, successfully solved the long-standing problem of protein folding, which had hindered progress in biology and medicine for decades.”

    From Ethical Framework for Harnessing the Power of AI in Healthcare and Beyond (2024) · cites this paper as [144]

  • The potential applications of AI in healthcare are vast, including rapid diagnosis, personalized care, and reduction of unnecessary outpatient visits, resulting in significant cost savings.

    “The potential applications of AI in healthcare are vast, including rapid diagnosis, personalized care, and reduction of unnecessary outpatient visits, resulting in significant cost savings [144].”

    From Ethical Framework for Harnessing the Power of AI in Healthcare and Beyond (2024) · cites this paper as [144]

  • AlphaFold is a software platform that uses protein sequence data and AI to predict proteins' corresponding three-dimensional structures.

    “It is a powerful algorithm that uses protein sequence data and AI to predict the proteins’ corresponding three-dimensional structures.”

    From The role of AI in drug discovery: challenges, opportunities, and strategies (2023) · cites this paper as [19]

  • AlphaFold2 in its by-default form was not optimized to sample conformations beyond the native structure.

    “even though in its by-default form AlphaFold2 was not optimized to sample conformations beyond the native structure,265”

    From Advancing Targeted Protein Degradation via Multiomics Profiling and Artificial Intelligence (2023) · cites this paper as [265]

Cites

Flagged first, then by citations

Research works printed but not linked: 5

  1. 10.1101/2021.1111.1121.469457 · DOI
  2. Christensen N. G., 1989, Tidsskr Nor Laegeforen, 109, 1770
  3. Bosshard B., 1983, Morphol Med., 3, 71
  4. 10.1101/2020.1107.1128.225490 · DOI
  5. Ingraham, J.; Riesselman, A.; Sander, C.; Marks, D. Learning protein structure with a differentiable simulator. https://openreview.net/forum?id=Byg3y3C9Km (accessed Sep 27, 2019).

Cited by

Sorted by citations

Related

Authors
Nussinov, R., Zhang, M., Liu, Y., & Jang, H.
Affiliations
  • Tel Aviv University (IL)
  • Frederick National Laboratory for Cancer Research (US)
Corresponding author
Nussinov, Ruth
ORCID
Venue
The Journal of Physical Chemistry B
Publisher
American Chemical Society (ACS)
Published
2022-08-17
Volume
126
Issue
34
Pages
6372-6383
ISSN
  • 1520-5207 (ISSN-L)
  • 1520-6106
DOI
10.1021/acs.jpcb.2c04346
PMC
PMC9442638
OpenAlex
W4292112379
Cited by
190
References
205
Citations by year
2026: 40 · 2025: 55 · 2024: 40 · 2023: 49 · 2022: 6
FWCI
3.90
Type
journal article
Language
en
Open access
Licence
Locations
  • The Journal of Physical Chemistry B (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
Funders
National Cancer Institute (HHSN261201500003I)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
Keywords
AlphaFold, protein structure prediction, allosteric regulation, protein folding, conformational ensembles, intrinsically disordered proteins, protein-protein interaction, drug discovery, deep learning, molecular dynamics simulation, microbiota-host interactions
MeSH terms
Allosteric Regulation, Artificial Intelligence, Humans, Protein Conformation, Protein Folding, Molecular Dynamics Simulation, Intrinsically Disordered Proteins

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 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

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