Flowr.root - A flow matching based foundation model for joint multi-purpose structure-aware 3D ligand generation and...

Area of Science:

• Computational chemistry and cheminformatics
• Artificial intelligence in drug discovery
• Molecular modeling and simulation

Background:

• Drug discovery relies on generating novel molecules with desired properties.
• Predicting binding affinity is crucial for identifying effective drug candidates.
• Existing methods often lack accuracy or efficiency in molecular generation and affinity prediction.

Purpose of the Study:

• To introduce Flowr.root, an SE(3)-equivariant flow-matching model for pocket-aware 3D ligand generation.
• To enable joint binding affinity prediction and confidence estimation.
• To support multiple molecular design modes and multi-endpoint affinity predictions.

Main Methods:

• Utilized SE(3)-equivariant flow-matching for 3D ligand generation.
• Trained on large-scale ligand libraries and mixed-fidelity protein-ligand complexes.
• Refined models on curated co-crystal datasets and employed parameter-efficient finetuning.
• Integrated an affinity prediction module for enhanced accuracy.

Main Results:

• Achieved state-of-the-art performance in unconditional and pocket-conditional 3D ligand generation.
• Demonstrated superior accuracy in binding affinity prediction on benchmark datasets.
• Showcased computational efficiency and geometric realism in generated structures.
• Validated through case studies with significant correlation between predicted and experimental binding energies.

Conclusions:

• Flowr.root provides a unified framework for structure-based drug design.
• The model accelerates drug discovery from hit identification to lead optimization.
• Continuous finetuning enables adaptation to project-specific structure-activity landscapes.
• Joint generation and affinity prediction capabilities steer molecular design effectively.