PocketGen: Enhancing Protein-Ligand Binding through Advanced Computational Design

Designing small-molecule-binding proteins, such as enzymes and biosensors, is essential in protein biology and bioengineering. Generating high-fidelity protein pockets—areas where proteins interact with ligand molecules—is challenging due to the complex interactions between ligand molecules and proteins, the flexibility of ligand molecules and amino acid side chains, and intricate sequence-structure dependencies. PocketGen is a novel deep generative method that addresses these challenges by producing the residue sequence and the full-atom structure within the protein pocket region, leveraging sequence-structure consistency.

Background

Protein-ligand interactions are fundamental to many biological processes, including enzymatic catalysis, signal transduction, and regulatory mechanisms within cells. Traditional methods for designing ligand-binding protein pockets, such as physics-based modeling and template-matching, face significant challenges due to the complexity of these interactions. These methods are often time-consuming and may not always achieve the desired level of precision in modeling protein-ligand interactions.

Traditional Methods

Physics-Based Methods
Physics-based methods, such as PocketOptimizer, develop pipelines that predict mutations in protein pockets to increase binding affinity based on physics-inspired energy functions and search algorithms. These methods often require extensive computational resources and time to generate high-affinity protein pockets.

Template-Matching Methods
Template-matching methods, like those employed by Polizzi et al., use a two-step strategy for pocket design. They first identify and assemble disconnected protein motifs surrounding the target molecule to build protein-ligand interactions. These motifs are then grafted onto the protein scaffold, and the best combinations of protein-ligand pairs are selected using scoring functions.

PocketGen Overview

PocketGen represents a significant advancement in protein pocket design. It integrates a bilevel graph transformer for structural encoding and a sequence refinement module utilizing a protein language model (pLM) for sequence prediction. This approach allows PocketGen to generate high-fidelity protein pockets with higher binding affinity and validity than state-of-the-art methods.

PocketGen Architecture

Bilevel Graph Transformer
The bilevel graph transformer captures interactions at multiple granularities (atom-level and residue/ligand-level) and aspects (intra-protein and protein-ligand) through bilevel attention mechanisms. This module ensures that the generated protein pockets maintain structural consistency with the ligand molecules they bind.

Sequence Refinement Module
The sequence refinement module employs a structural adapter integrated into the pLM for sequence updates. This adapter facilitates cross-attention between sequence and structure features, promoting information flow and achieving sequence-structure consistency. Only the adapter is fine-tuned during training, while the other layers of the pLM remain unchanged.

Experimental Methods

Datasets Used
PocketGen was trained and evaluated on two widely used datasets: the CrossDocked dataset and the Binding MOAD dataset. These datasets comprise protein-ligand pairs generated through cross-docking and experimentally determined protein-ligand complexes, respectively.

Training Setup
During training, only the structural adapter in the pLM is fine-tuned. The model's performance is evaluated based on its ability to generate protein pockets with high binding affinity and structural validity.

Evaluation Metrics
The quality of generated protein pockets is assessed using several metrics, including the AutoDock Vina score, scRMSD, scTM, and pLDDT. These metrics evaluate the affinity between the generated pocket and the target ligand molecule, the structural validity of the generated pockets, and the confidence in structural predictions.

Results

Benchmarking Results
PocketGen outperforms existing methods in protein pocket generation across multiple benchmarks. It achieves higher binding affinity and structural validity than state-of-the-art methods, with a success rate of 97% in creating pockets with higher affinity than reference cases.

Binding Affinity
The dissociation constants determined through binding assays confirm PocketGen's strong affinity for ligand molecules. The generated pockets consistently exhibit higher binding affinity compared to those produced by traditional methods.

Structural Validity
PocketGen maintains structural consistency across protein sequence and structure domains, achieving high scRMSD, scTM, and pLDDT scores. This consistency is critical for ensuring the functional stability of the designed protein pockets.

Case Studies

Cortisol-Binding Antibody (HCY)
PocketGen successfully redesigned the pocket of a cortisol-specific antibody, enhancing its binding affinity and stability. The redesigned pocket maintained essential interaction patterns while introducing additional hydrogen bond-mediated interactions.

Apixaban (APX)
For apixaban, an oral anticoagulant, PocketGen redesigned the pocket of Factor Xa, a crucial enzyme in blood coagulation. The generated pocket established additional interactions, such as π-Cation interactions, enhancing its therapeutic potential.

Fentanyl-Binding Proteins (7v7)
PocketGen generated high-affinity pockets for fentanyl, a widely abused drug. The redesigned pockets established additional non-covalent interactions, enhancing their potential for use in biosensors for detecting and neutralizing the toxin.

Discussion

Comparison with Existing Methods
PocketGen's ability to capture interactions at multiple granularities and aspects sets it apart from traditional methods. Its integration of pLMs for sequence refinement ensures sequence-structure consistency, leading to superior performance in generating high-affinity protein pockets.

Biological Significance
The discovery of PocketGen's dual binding abilities opens new avenues for understanding the functions of ligand-binding proteins. Its potential applications in drug discovery and protein engineering highlight its significance in the field of molecular biology.

Future Directions

Enhancing Scalability
Future research will focus on enhancing PocketGen's scalability to design larger areas of the protein beyond the pocket region. This will require modifications to improve its robustness and performance in generating larger protein structures.

Incorporating Additional Biochemical Priors
Incorporating additional biochemical priors, such as subpockets and interaction templates, will improve PocketGen's generalizability and success rates. These enhancements will enable more precise and efficient protein pocket design.

Conclusion

In summary, PocketGen represents a significant advancement in protein pocket design, addressing the limitations of traditional methods. Its ability to generate high-affinity, structurally valid protein pockets with dual binding abilities highlights its potential in protein engineering and drug discovery. Future research will further explore its applications and enhance its scalability and robustness.

FAQs

Q1. What is PocketGen?
PocketGen is a deep generative method for designing high-fidelity ligand-binding protein pockets by producing the residue sequence and full-atom structure within the pocket region.

Q2. How does PocketGen differ from traditional methods?
Unlike traditional methods, PocketGen integrates a bilevel graph transformer and a sequence refinement module using protein language models to ensure sequence-structure consistency and higher binding affinity.

Q3. What datasets were used to train and evaluate PocketGen?
PocketGen was trained and evaluated on the CrossDocked and Binding MOAD datasets, which include protein-ligand pairs generated through cross-docking and experimentally determined protein-ligand complexes.

Q4. What metrics are used to evaluate PocketGen's performance?
PocketGen's performance is evaluated using metrics such as the AutoDock Vina score, scRMSD, scTM, and pLDDT, which assess binding affinity, structural validity, and confidence in structural predictions.

Q5. What are the future directions for PocketGen research?
Future research will focus on enhancing PocketGen's scalability to design larger protein areas and incorporating additional biochemical priors to improve generalizability and success rates.