Pharmolix-FM: All-Atom Foundation Models for Molecular Modeling and Generation

2025-03-12

AI Summary

Current all-atom foundation models exhibit limited generalization in molecular modeling and generation, primarily due to the multimodal nature of atomic data and the absence of systematic design in training and sampling strategies. To address this, we propose PharMolixFM—the first unified foundation model framework for all-atom structure prediction, docking, and generation of both biomacromolecules and small molecules. Our approach innovatively integrates multimodal generative architectures (diffusion, autoregressive, and normalizing flow), introduces a task-prior-guided generalized denoising paradigm, and empirically identifies a scaling law governing reasoning efficiency in molecular generation. On protein–ligand docking, PharMolixFM-Diff achieves 83.9% accuracy (RMSD < 2 Å) given a binding pocket, with substantially accelerated inference. The code and pre-trained models are publicly released.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Generation
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Structural biology relies on accurate three-dimensional biomolecular structures to advance our understanding of biological functions, disease mechanisms, and therapeutics. While recent advances in deep learning have enabled the development of all-atom foundation models for molecular modeling and generation, existing approaches face challenges in generalization due to the multi-modal nature of atomic data and the lack of comprehensive analysis of training and sampling strategies. To address these limitations, we propose PharMolixFM, a unified framework for constructing all-atom foundation models based on multi-modal generative techniques. Our framework includes three variants using state-of-the-art multi-modal generative models. By formulating molecular tasks as a generalized denoising process with task-specific priors, PharMolixFM achieves robust performance across various structural biology applications. Experimental results demonstrate that PharMolixFM-Diff achieves competitive prediction accuracy in protein-small-molecule docking (83.9% vs. 90.2% RMSD<2{AA}, given pocket) with significantly improved inference speed. Moreover, we explore the empirical inference scaling law by introducing more sampling repeats or steps. Our code and model are available at https://github.com/PharMolix/OpenBioMed.

Problem

• Addresses generalization challenges in all-atom molecular modeling
• Improves multi-modal generative techniques for structural biology
• Enhances protein-small-molecule docking accuracy and speed

Innovation

• Multi-modal generative techniques for molecular modeling
• Generalized denoising process with task-specific priors
• Improved inference speed and prediction accuracy