Blind protein-ligand docking with diffusion-based deep generative models

Corso, Gabriele; Jing, Bowen; Stark, Hannes; Barzilay, Regina; Jaakkola, Tommi S.

doi:10.1016/j.bpj.2022.11.937

Cited by 100 publications

(160 citation statements)

References 0 publications

Supporting

Mentioning

124

Contrasting

Order By: Relevance

“…Most famously, the AI/ML algorithm AlphaFold 2 (Jumper et al, 2021) (and to a lesser extent RoseTTAfold (Baek et al, 2021)) made a quantum leap in protein structure prediction accuracy. More relevant to the work reported here, AI/ML is being used to great effect for structure-based drug design and computational chemistry, including protein-ligand docking (Corso et al, 2022) and ligand design (Wallach et al, 2015). Importantly, all of these methods rely on training data in the form of experimental protein structures from the PDB, the vast majority of which are cryo-temperature crystal structures.…”

Section: Discussionmentioning

confidence: 99%

Room-temperature crystallography reveals altered binding of small-molecule fragments to PTP1B

Mehlman

Biel

Azeem

et al. 2023

eLife

View full text Add to dashboard Cite

Much of our current understanding of how small-molecule ligands interact with proteins stems from X-ray crystal structures determined at cryogenic (cryo) temperature. For proteins alone, room-temperature (RT) crystallography can reveal previously hidden, biologically relevant alternate conformations. However, less is understood about how RT crystallography may impact the conformational landscapes of protein-ligand complexes. Previously, we showed that small-molecule fragments cluster in putative allosteric sites using a cryo crystallographic screen of the therapeutic target PTP1B (Keedy et al., 2018). Here, we have performed two RT crystallographic screens of PTP1B using many of the same fragments, representing the largest RT crystallographic screens of a diverse library of ligands to date, and enabling a direct interrogation of the effect of data collection temperature on protein-ligand interactions. We show that at RT, fewer ligands bind, and often more weakly – but with a variety of temperature-dependent differences, including unique binding poses, changes in solvation, new binding sites, and distinct protein allosteric conformational responses. Overall, this work suggests that the vast body of existing cryo-temperature protein-ligand structures may provide an incomplete picture, and highlights the potential of RT crystallography to help complete this picture by revealing distinct conformational modes of protein-ligand systems. Our results may inspire future use of RT crystallography to interrogate the roles of protein-ligand conformational ensembles in biological function.

show abstract

Section: Discussionmentioning

confidence: 99%

Room-temperature crystallography reveals altered binding of small-molecule fragments to PTP1B

Mehlman

Biel

Azeem

et al. 2023

eLife

View full text Add to dashboard Cite

show abstract

“…We and others have shown that this issue can be mitigated by sampling multiple poses and re-ranking them with a deep learning model ( Stafford et al, 2022 ), often yielding sufficiently accurate poses for applications such as the one included in this manuscript. Recent advances in molecular docking ( Corso et al, 2022 ) have led to docking tools that do away with explicitly defined docking scoring functions. If these tools could be run at the scale needed to generate training data for the AtomNet pKi predictor, it is possible that the model could achieve better predictive performance.…”

Section: Discussionmentioning

confidence: 99%

Discovery of a cryptic pocket in the AI-predicted structure of PPM1D phosphatase explains the binding site and potency of its allosteric inhibitors

Meller

Oliveira

Davtyan

et al. 2023

Front. Mol. Biosci.

View full text Add to dashboard Cite

Virtual screening is a widely used tool for drug discovery, but its predictive power can vary dramatically depending on how much structural data is available. In the best case, crystal structures of a ligand-bound protein can help find more potent ligands. However, virtual screens tend to be less predictive when only ligand-free crystal structures are available, and even less predictive if a homology model or other predicted structure must be used. Here, we explore the possibility that this situation can be improved by better accounting for protein dynamics, as simulations started from a single structure have a reasonable chance of sampling nearby structures that are more compatible with ligand binding. As a specific example, we consider the cancer drug target PPM1D/Wip1 phosphatase, a protein that lacks crystal structures. High-throughput screens have led to the discovery of several allosteric inhibitors of PPM1D, but their binding mode remains unknown. To enable further drug discovery efforts, we assessed the predictive power of an AlphaFold-predicted structure of PPM1D and a Markov state model (MSM) built from molecular dynamics simulations initiated from that structure. Our simulations reveal a cryptic pocket at the interface between two important structural elements, the flap and hinge regions. Using deep learning to predict the pose quality of each docked compound for the active site and cryptic pocket suggests that the inhibitors strongly prefer binding to the cryptic pocket, consistent with their allosteric effect. The predicted affinities for the dynamically uncovered cryptic pocket also recapitulate the relative potencies of the compounds (τb = 0.70) better than the predicted affinities for the static AlphaFold-predicted structure (τb = 0.42). Taken together, these results suggest that targeting the cryptic pocket is a good strategy for drugging PPM1D and, more generally, that conformations selected from simulation can improve virtual screening when limited structural data is available.

show abstract

“…This strategy was shown to outperform previous traditional and DL docking protocols. 75 Meller and co-workers also developed an AF2-based strategy to find cryptic pockets. 76 DL has also been applied for finding potential location sites of transition metals in proteins (Metal1D and Metal3D).…”

Section: Application Of Af2 and Other Deep Learning Techniques For Pr...mentioning

confidence: 99%

AlphaFold2 and Deep Learning for Elucidating Enzyme Conformational Flexibility and Its Application for Design

Casadevall

Duran

Osuna

2023

JACS Au

View full text Add to dashboard Cite

The recent success of AlphaFold2 (AF2) and other deep learning (DL) tools in accurately predicting the folded three-dimensional (3D) structure of proteins and enzymes has revolutionized the structural biology and protein design fields. The 3D structure indeed reveals key information on the arrangement of the catalytic machinery of enzymes and which structural elements gate the active site pocket. However, comprehending enzymatic activity requires a detailed knowledge of the chemical steps involved along the catalytic cycle and the exploration of the multiple thermally accessible conformations that enzymes adopt when in solution. In this Perspective, some of the recent studies showing the potential of AF2 in elucidating the conformational landscape of enzymes are provided. Selected examples of the key developments of AF2-based and DL methods for protein design are discussed, as well as a few enzyme design cases. These studies show the potential of AF2 and DL for allowing the routine computational design of efficient enzymes.

show abstract

Blind protein-ligand docking with diffusion-based deep generative models

Cited by 100 publications

References 0 publications

Room-temperature crystallography reveals altered binding of small-molecule fragments to PTP1B

Room-temperature crystallography reveals altered binding of small-molecule fragments to PTP1B

Discovery of a cryptic pocket in the AI-predicted structure of PPM1D phosphatase explains the binding site and potency of its allosteric inhibitors

AlphaFold2 and Deep Learning for Elucidating Enzyme Conformational Flexibility and Its Application for Design

Contact Info

Product

Resources

About