Modeling Boltzmann-weighted structural ensembles of proteins using artificial intelligence–based methods

Current Opinion in Structural Biology, Год журнала: 2025, Номер 91, С. 103000 - 103000

Опубликована: Фев. 8, 2025

Язык: Английский

---

Adaptive Sampling Methods for Molecular Dynamics in the Era of Machine Learning

The Journal of Physical Chemistry B, Год журнала: 2023, Номер 127(50), С. 10669 - 10681

Опубликована: Дек. 11, 2023

Molecular dynamics (MD) simulations are fundamental computational tools for the study of proteins and their free energy landscapes. However, sampling protein conformational changes through MD is challenging due to relatively long time scales these processes. Many enhanced approaches have emerged tackle this problem, including biased path-sampling methods. In Perspective, we focus on adaptive algorithms. These techniques differ from other because thermodynamic ensemble preserved solely by restarting trajectories at particularly chosen seeds rather than introducing biasing forces. We begin our treatment with an overview theoretically transparent methods, where discuss principles guidelines sampling. Then, present a brief summary select methods that been applied realistic systems in past. Finally, recent advances methodology powered deep learning techniques, as well shortcomings.

Язык: Английский

---

The role of artificial intelligence in drug screening, drug design, and clinical trials

Frontiers in Pharmacology, Год журнала: 2024, Номер 15

Опубликована: Ноя. 29, 2024

The role of computational tools in drug discovery and development is becoming increasingly important due to the rapid computing power advancements chemistry biology, improving research efficiency reducing costs potential risks preclinical clinical trials. Machine learning, especially deep a subfield artificial intelligence (AI), has demonstrated significant advantages development, including high-throughput virtual screening,

Язык: Английский

---

AlphaFold and Protein Folding: Not Dead Yet! The Frontier Is Conformational Ensembles

Annual Review of Biomedical Data Science, Год журнала: 2024, Номер 7(1), С. 51 - 57

Опубликована: Апрель 11, 2024

Like the black knight in classic Monty Python movie, grand scientific challenges such as protein folding are hard to finish off. Notably, AlphaFold is revolutionizing structural biology by bringing highly accurate structure prediction masses and opening up innumerable new avenues of research. Despite this enormous success, calling prediction, much less related problems, “solved” dangerous, doing so could stymie further progress. Imagine what world would be like if we had declared flight solved after first commercial airlines opened stopped investing research development. Likewise, there still important limitations that benefit from addressing. Moreover, limited our understanding diversity different structures a single can adopt (called conformational ensemble) dynamics which explores space. What clear ensembles critical function, aspect will advance ability design proteins drugs.

Язык: Английский

---

Molecular dynamics simulations for the structure-based drug design: targeting small-GTPases proteins

Expert Opinion on Drug Discovery, Год журнала: 2024, Номер 19(10), С. 1259 - 1279

Опубликована: Авг. 6, 2024

Molecular Dynamics (MD) simulations can support mechanism-based drug design. Indeed, MD by capturing biomolecule motions at finite temperatures reveal hidden binding sites, accurately predict drug-binding poses, and estimate the thermodynamics kinetics, crucial information for discovery campaigns. Small-Guanosine Triphosphate Phosphohydrolases (GTPases) regulate a cascade of signaling events, that affect most cellular processes. Their deregulation is linked to several diseases, making them appealing targets. The broad roles small-GTPases in processes recent approval covalent KRas inhibitor as an anticancer agent renewed interest targeting small-GTPase with small molecules.

Язык: Английский

---

Empowering AlphaFold2 for protein conformation selective drug discovery with AlphaFold2-RAVE

Опубликована: Авг. 14, 2024

Small molecule drug design hinges on obtaining co-crystallized ligand-protein structures. Despite AlphaFold2’s strides in protein native structure prediction, its focus apo structures overlooks ligands and associated holo Moreover, designing selective drugs often benefits from the targeting of diverse metastable conformations. Therefore, direct application AlphaFold2 models virtual screening dis-covery remains tentative. Here, we demonstrate an based framework combined with all-atom enhanced sampling molecular dynamics induced fit docking, named AF2RAVE-Glide, to conduct computational model small binding kinase conformations, initiated sequences. We AF2RAVE-Glide workflow three different kinases their type I II inhibitors, special emphasis known inhibitors which target classical DFG-out state. These states are not easy sample AlphaFold2. Here how AF2RAVE these conformations can be sampled for high enough ac- curacy enable subsequent docking more than 50% success rates across calculations. believe protocol should deployable other proteins generally.

Язык: Английский

---

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

Frontiers in Molecular Biosciences, Год журнала: 2023, Номер 10

Опубликована: Апрель 18, 2023

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

Язык: Английский

---

A journey from molecule to physiology and in silico tools for drug discovery targeting the transient receptor potential vanilloid type 1 (TRPV1) channel

Frontiers in Pharmacology, Год журнала: 2024, Номер 14

Опубликована: Янв. 24, 2024

The heat and capsaicin receptor TRPV1 channel is widely expressed in nerve terminals of dorsal root ganglia (DRGs) trigeminal innervating the body face, respectively, as well other tissues organs including central nervous system. a versatile that detects harmful heat, pain, various internal external ligands. Hence, it operates polymodal sensory channel. Many pathological conditions neuroinflammation, cancer, psychiatric disorders, are linked to abnormal functioning peripheral tissues. Intense biomedical research underway discover compounds can modulate provide pain relief. molecular mechanisms underlying temperature sensing remain largely unknown, although they closely transduction. Prolonged exposure generates analgesia, hence numerous analogs have been developed efficient analgesics for emergence silico tools offered significant techniques modeling machine learning algorithms indentify druggable sites repositioning current drugs aimed at TRPV1. Here we recapitulate physiological pathophysiological functions channel, structural models obtained through cryo-EM, pharmacological tested on TRPV1, drug discovery repositioning.

Язык: Английский

---

Empowering AlphaFold2 for protein conformation selective drug discovery with AlphaFold2-RAVE

eLife, Год журнала: 2024, Номер 13

Опубликована: Июль 1, 2024

Small molecule drug design hinges on obtaining co-crystallized ligand-protein structures. Despite AlphaFold2's strides in protein native structure prediction, its focus apo structures overlooks ligands and associated holo Moreover, designing selective drugs often benefits from the targeting of diverse metastable conformations. Therefore, direct application AlphaFold2 models virtual screening discovery remains tentative. Here, we demonstrate an based framework combined with all-atom enhanced sampling molecular dynamics induced fit docking, named AF2RAVE-Glide, to conduct computational model small binding kinase conformations, initiated sequences. We AF2RAVE-Glide workflow three different kinases their type I II inhibitors, special emphasis known inhibitors which target classical DFG-out state. These states are not easy sample AlphaFold2. Here how AF2RAVE these conformations can be sampled for high enough accuracy enable subsequent docking more than 50% success rates across calculations. believe protocol should deployable other proteins generally.

Язык: Английский

---

APACE: AlphaFold2 and advanced computing as a service for accelerated discovery in biophysics

Proceedings of the National Academy of Sciences, Год журнала: 2024, Номер 121(27)

Опубликована: Июнь 24, 2024

The prediction of protein 3D structure from amino acid sequence is a computational grand challenge in biophysics and plays key role robust algorithms, drug discovery to genome interpretation. advent AI models, such as AlphaFold, revolutionizing applications that depend on algorithms. To maximize the impact, ease usability, these tools we introduce APACE, AlphaFold2 advanced computing service, framework effectively handles this model its TB-size database conduct accelerated analyses modern supercomputing environments. We deployed APACE Delta Polaris supercomputers quantified performance for accurate predictions using four exemplar proteins: 6AWO, 6OAN, 7MEZ, 6D6U. Using up 300 ensembles, distributed across 200 NVIDIA A100 GPUs, found two orders magnitude faster than off-the-self implementations, reducing time-to-solution weeks minutes. This approach may be readily linked with robotics laboratories automate accelerate scientific discovery.

Язык: Английский

---