Empowering AlphaFold2 for protein conformation selective drug discovery with AlphaFold2-RAVE (2404.07102v3)

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

• The paper introduces AF2RAVE-Glide, a novel framework combining AlphaFold2 with enhanced sampling and induced fit docking to identify and exploit metastable protein conformations crucial for selective drug discovery.
• Using this approach, researchers successfully sampled conformations like DFG-out states and achieved over 50% success rates in docking type II inhibitors to kinases (Abl1, DDR1, Src), a task challenging for standard AlphaFold2 models.
• This work demonstrates a significant step in integrating AI-based protein structure prediction with advanced simulation techniques, potentially expanding computational drug discovery to diverse protein classes and future generative AI applications.

Overview of Empowering AlphaFold2 for Protein Conformation Selective Drug Discovery with AlphaFold2-RAVE

The paper "Empowering AlphaFold2 for protein conformation selective drug discovery with AlphaFold2-RAVE" investigates the integration of advanced molecular modeling techniques with AlphaFold2 (AF2) to address challenges in structure-based drug design. The authors propose a framework named AF2RAVE-Glide, which synergistically combines AF2 for protein structure prediction with enhanced sampling techniques and induced fit docking (IFD) to explore and exploit metastable protein conformations that are not readily accessible through traditional AF2 approaches.

Scientific Context and Challenges

Liquid molecule drug design heavily relies on accurate structures of ligand-protein complexes, often obtained via co-crystallization. AlphaFold2 has significantly advanced the prediction of native protein (apo) structures; however, it does not target the ligand-bound states (holo structures). Effective drug discovery, particularly for kinase inhibitors, requires targeting diverse metastable conformations that are often overlooked in conventional structure prediction methods. This paper addresses the gap by leveraging AF2-derived models for sampling protein metastable states and suggests enriching these with suitable ligand-binding conformations using a novel protocol.

AF2RAVE-Glide Workflow

The paper introduces the AF2RAVE-Glide protocol, which begins with AF2 to generate putative protein conformations from sequences. These conformations serve as initial models for Reweighted Autoencoded Variational Bayes for Enhanced Sampling (RAVE) to explore metastable states. The AF2RAVE method incorporates reduced multiple sequence alignment (rMSA AF2) and latent space sampling, allowing the identification of metastable states and the computation of associated free energy profiles. Structural candidates that potentially host ligand-binding sites are subsequently subjected to induced fit docking through Glide XP and IFD, with the goal of refining and validating ligand-binding pockets.

Numerical Results

In a proof-of-concept paper, the authors focus on three kinase proteins—Abl1, DDR1, and Src—and their interactions with type I and II inhibitors. The AF2RAVE-Glide approach excels at identifying metastable conformations suitable for type II inhibitor docking, achieving ligand binding poses with success rates greater than 50% across different docking scenarios. Notably, this methodology potentates sampling of DFG-out states, which are critical for type II inhibitors binding, a task at which standard AF2 models falter. Such synergy highlights the latent space exploration capability of AF2RAVE and its potential in expanding the application scope of AF2 models.

Implications and Future Directions

The proposed AF2RAVE-Glide framework represents a substantive advancement in computational drug discovery. It marks a shift towards integrating modern AI-based protein models with sophisticated molecular dynamics simulations and enhanced sampling techniques. This integration not only enriches the conformational diversity necessary for ligand-binding site prediction but also significantly enhances docking accuracy and success.

The paper suggests future developments in expanding this approach to other protein classes, notably G-protein-coupled receptors, due to their pharmacological importance. Furthermore, it hints at the incorporation of generative AI algorithms to hypothesize conformation states beyond those sampled, provided these predictions are backed with stringent free energy validations. Such innovations could usher in a new era of AI-driven structural biology, allowing rapid hypothesis testing and verification against large chemical libraries.

Conclusion

Overall, the integration of enhanced sampling with AlphaFold2 as outlined in this paper has the potential to profoundly impact drug discovery processes, providing a blueprint for future endeavors to incorporate AI insights with physical method validations in structural biology. The AF2RAVE-Glide framework thus exemplifies a significant step towards refining computational approaches for selective drug discovery, leveraging the untapped potential of metastable conformations in kinase proteins and beyond.