How accurately can one predict drug binding modes using AlphaFold models?

eLife, Год журнала: 2023, Номер 12

Опубликована: Дек. 22, 2023

Computational prediction of protein structure has been pursued intensely for decades, motivated largely by the goal using structural models drug discovery. Recently developed machine-learning methods such as AlphaFold 2 (AF2) have dramatically improved prediction, with reported accuracy approaching that experimentally determined structures. To what extent do these advances translate to an ability predict more accurately how drugs and candidates bind their target proteins? Here, we carefully examine utility AF2 predicting binding poses drug-like molecules at largest class targets, G-protein-coupled receptors. We find capture pocket structures much than traditional homology models, errors nearly small differences between same different ligands bound. Strikingly, however, ligand-binding predicted computational docking is not significantly higher when lower without These results important implications all those who might use

Язык: Английский

---

The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins

Nature Chemical Biology, Год журнала: 2024, Номер 20(8), С. 950 - 959

Опубликована: Июнь 21, 2024

Язык: Английский

---

AlphaFold prediction of structural ensembles of disordered proteins

Nature Communications, Год журнала: 2025, Номер 16(1)

Опубликована: Фев. 14, 2025

Abstract Deep learning methods of predicting protein structures have reached an accuracy comparable to that high-resolution experimental methods. It is thus possible generate accurate models the native states hundreds millions proteins. An open question, however, concerns whether these advances can be translated disordered proteins, which should represented as structural ensembles because their heterogeneous and dynamical nature. To address this problem, we introduce AlphaFold-Metainference method use AlphaFold-derived distances restraints in molecular dynamics simulations construct ordered The results obtained using illustrate possibility making predictions conformational properties proteins deep trained on large databases available for folded

Язык: Английский

---

Evaluation of AlphaFold2 structures as docking targets

Protein Science, Год журнала: 2022, Номер 32(1)

Опубликована: Дек. 7, 2022

AlphaFold2 is a promising new tool for researchers to predict protein structures and generate high-quality models, with low backbone global root-mean-square deviation (RMSD) when compared experimental structures. However, it unclear if the predicted by will be valuable targets of docking. To address this question, we redocked ligands in PDBbind datasets against co-crystallized receptor using AutoDock-GPU. We find that quality measure provided during structure prediction not good predictor docking performance, despite accurately reflecting alpha carbon alignment Removing low-confidence regions making side chains flexible improves outcomes. Overall, conformation, fine structural details limit naive application models as targets.

Язык: Английский

---

Are Deep Learning Structural Models Sufficiently Accurate for Free-Energy Calculations? Application of FEP+ to AlphaFold2-Predicted Structures

Journal of Chemical Information and Modeling, Год журнала: 2022, Номер 62(18), С. 4351 - 4360

Опубликована: Сен. 13, 2022

The availability of AlphaFold2 has led to great excitement in the scientific community─particularly among drug hunters─due ability algorithm predict protein structures with high accuracy. However, beyond globally accurate structure prediction, it remains be determined whether ligand binding sites are predicted sufficient accuracy these useful supporting computationally driven discovery programs. We explored this question by performing free-energy perturbation (FEP) calculations on a set well-studied protein–ligand complexes, where predictions were performed removing all templates >30% identity target from training set. observed that most cases, ΔΔG values for transformations calculated FEP, using prospective structures, comparable corresponding previously carried out crystal structures. conclude under right circumstances, AlphaFold2-modeled enough used physics-based methods such as FEP typical lead optimization stages program.

Язык: Английский

---

How accurately can one predict drug binding modes using AlphaFold models?

eLife, Год журнала: 2023, Номер 12

Опубликована: Авг. 8, 2023

Computational prediction of protein structure has been pursued intensely for decades, motivated largely by the goal using structural models drug discovery. Recently developed machine-learning methods such as AlphaFold 2 (AF2) have dramatically improved prediction, with reported accuracy approaching that experimentally determined structures. To what extent do these advances translate to an ability predict more accurately how drugs and candidates bind their target proteins? Here, we carefully examine utility AF2 predicting binding poses drug-like molecules at largest class targets, G-protein-coupled receptors. We find capture pocket structures much than traditional homology models, errors nearly small differences between same different ligands bound. Strikingly, however, ligand-binding predicted computational docking is not significantly higher when lower without These results important implications all those who might use

Язык: Английский

---

Recognition of methamphetamine and other amines by trace amine receptor TAAR1

Nature, Год журнала: 2023, Номер 624(7992), С. 663 - 671

Опубликована: Ноя. 7, 2023

Язык: Английский

---

AlphaFold2 models of the active form of all 437 catalytically competent human protein kinase domains

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Июль 25, 2023

Humans have 437 catalytically competent protein kinase domains with the typical fold, similar to structure of Protein Kinase A (PKA). Only 155 these kinases are in Data Bank their active form. The form a must satisfy requirements for binding ATP, magnesium, and substrate. From structural bioinformatics analysis 40 unique substrate-bound kinases, we derived several criteria kinases. We include on DFG motif activation loop but also positions N-terminal C-terminal segments that be placed appropriately bind Because catalytic is needed understanding substrate specificity effects mutations activity cancer other diseases, used AlphaFold2 produce models all human This was accomplished templates from PDB shallow multiple sequence alignments orthologs close homologs query protein. selected each based pLDDT scores residues, demonstrating highest scoring lowest or RMSD 22 non-redundant structures PDB. larger benchmark 130 complete loops shows 80% highest-scoring < 1.0 Å 90% 2.0 over backbone atoms. Models available at http://dunbrack.fccc.edu/kincore/activemodels. believe they may useful interpreting leading constitutive as well modeling inhibitor molecules which state.

Язык: Английский

---

From byte to bench to bedside: molecular dynamics simulations and drug discovery

BMC Biology, Год журнала: 2023, Номер 21(1)

Опубликована: Дек. 29, 2023

Molecular dynamics (MD) simulations and computer-aided drug design (CADD) have advanced substantially over the past two decades, thanks to continuous computer hardware software improvements. Given these advancements, MD are poised become even more powerful tools for investigating dynamic interactions between potential small-molecule drugs their target proteins, with significant implications pharmacological research.

Язык: Английский

---

Computational drug development for membrane protein targets

Nature Biotechnology, Год журнала: 2024, Номер 42(2), С. 229 - 242

Опубликована: Фев. 1, 2024

Язык: Английский

---