Abstract:Molecular docking, given a ligand molecule and a ligand binding site (called ``pocket'') on a protein, predicting the binding mode of the protein-ligand complex, is a widely used technique in drug design. Many deep learning models have been developed for molecular docking, while most existing deep learning models perform docking on the whole protein, rather than on a given pocket as the traditional molecular docking approaches, which does not match common needs. What's more, they claim to perform better than traditional molecular docking, but the approach of comparison is not fair, since traditional methods are not designed for docking on the whole protein without a given pocket. In this paper, we design a series of experiments to examine the actual performance of these deep learning models and traditional methods. For a fair comparison, we decompose the docking on the whole protein into two steps, pocket searching and docking on a given pocket, and build pipelines to evaluate traditional methods and deep learning methods respectively. We find that deep learning models are actually good at pocket searching, but traditional methods are better than deep learning models at docking on given pockets. Overall, our work explicitly reveals some potential problems in current deep learning models for molecular docking and provides several suggestions for future works.

What problem does this paper attempt to address?

The paper attempts to address issues primarily focused on the application of molecular docking techniques in drug design. Specifically, the paper focuses on the following aspects: 1. **Fair comparison between existing deep learning models and traditional methods**: Many existing deep learning models perform molecular docking directly on the entire protein (referred to as "blind docking") rather than targeting specific pockets as traditional methods do. This approach does not meet practical needs and unfairly compares traditional methods with deep learning models. 2. **The true performance of deep learning models**: Although some deep learning models claim to outperform traditional methods, the experiments and evaluation methods of these claims are questionable. The paper aims to fairly compare the performance of deep learning models and traditional methods by designing a series of experiments. 3. **Advantages and disadvantages of deep learning models**: The paper explores the performance of deep learning models in pocket search and molecular docking to clarify in which tasks they perform better. To answer these questions, the authors designed a series of experiments that decompose the molecular docking process into two steps: pocket search and molecular docking on a given pocket, and separately evaluated the performance of traditional methods and deep learning models in these two steps. Ultimately, the paper concludes the following: - Under the same pocket conditions, traditional molecular docking methods still outperform deep learning models. - Deep learning models excel in pocket search but are inferior to traditional methods in molecular docking on a given pocket. - Deep learning models have great potential, especially in pocket search, and may surpass traditional methods in the future. Through these studies, the paper reveals some potential issues with current deep learning models in molecular docking and provides suggestions for future related research.

Do Deep Learning Models Really Outperform Traditional Approaches in Molecular Docking?

Do Deep Learning Models Really Outperform Traditional Approaches in Molecular Docking?

Exploration on Learning Molecular Docking with Deep Learning Models

Deep Learning for Protein-Ligand Docking: Are We There Yet?

DeltaDock: A Unified Framework for Accurate, Efficient, and Physically Reliable Molecular Docking

Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows

Quantum-Inspired Machine Learning for Molecular Docking

Deep Learning Model of Dock by Dock Process Significantly Accelerate the Process of Docking-Based Virtual Screening

DiffDock: Diffusion Steps, Twists, and Turns for Molecular Docking

DOCKSTRING: Easy Molecular Docking Yields Better Benchmarks for Ligand Design

Deep learning is competing random forest in computational docking

Synergistic Application of Molecular Docking and Machine Learning for Improved Binding Pose

PoseBusters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences

Assessment of Protein-Protein Docking Models Using Deep Learning

DeepDock: Enhancing Ligand-protein Interaction Prediction by a Combination of Ligand and Structure Information

Deep Scoring Neural Network Replacing the Scoring Function Components to Improve the Performance of Structure-Based Molecular Docking.

Combining Docking Pose Rank and Structure with Deep Learning Improves Protein–Ligand Binding Mode Prediction over a Baseline Docking Approach

Pre-Training on Large-Scale Generated Docking Conformations with HelixDock to Unlock the Potential of Protein-ligand Structure Prediction Models

Protein–Ligand Docking in the Machine-Learning Era

Harnessing Deep Learning for Enhanced Ligand Docking.

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking