ChemFlow─From 2D Chemical Libraries to Protein-Ligand Binding Free Energies

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Accurate prediction of protein-ligand binding affinities is crucial for rational drug design.
• Existing computational methods are often too slow, inaccurate, or complex for widespread use in virtual screening.
• Sophisticated methods like molecular dynamics require extensive preprocessing, limiting accessibility for non-expert users.

Purpose of the Study:

• To develop a novel computational platform, ChemFlow, to streamline the prediction of protein-ligand binding affinities.
• To bridge the gap between 2D chemical libraries and reliable binding affinity estimations.
• To provide an accessible tool for preparing compound libraries, docking, and rescoring poses.

Main Methods:

• ChemFlow processes chemical libraries in SMILES or SDF formats.
• The platform performs molecular docking into specified protein binding sites.
• Simplified free energy calculations are used for pose rescoring, enhanced by GPU computing.

Main Results:

• ChemFlow demonstrated high accuracy, achieving an RMSE < 2 kcal/mol for relative binding free energies on a dataset of 626 protein-ligand complexes.
• The platform exhibits high throughput, capable of processing approximately 1000 ligands per day on a midsize computer cluster.
• The software integrates compound preparation, docking, and rescoring into a single workflow.

Conclusions:

• ChemFlow offers an efficient and accurate solution for predicting protein-ligand binding affinities.
• The platform democratizes computational drug discovery by simplifying complex workflows.
• ChemFlow is publicly available, facilitating its adoption in virtual high-throughput screening and drug design efforts.