CMADiff: Cross-Modal Aligned Diffusion for Controllable Protein Generation

2025-03-27

AI Summary

Existing AI-based protein design models predominantly rely on sequence or structural data, neglecting explicit modeling of physicochemical properties and lacking controllability under intuitive semantic conditions. To address this, we propose BioDiff—a novel framework that, for the first time, achieves three-modal alignment among protein physicochemical attributes, natural language descriptions, and amino acid sequences within a diffusion modeling paradigm. BioDiff integrates a conditional variational autoencoder (CVAE) with a conditional diffusion model and introduces BioAligner, a contrastive alignment module that jointly embeds textual semantics and quantitative physicochemical features—including hydrophobicity, net charge, and molecular weight. Generated proteins are rigorously validated using AlphaFold3 for structural plausibility and benchmarked across multiple metrics. Results demonstrate superior physicochemical validity and structural foldability compared to state-of-the-art methods. The implementation is publicly available.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

AI-assisted protein design has emerged as a critical tool for advancing biotechnology, as deep generative models have demonstrated their reliability in this domain. However, most existing models primarily utilize protein sequence or structural data for training, neglecting the physicochemical properties of proteins.Moreover, they are deficient to control the generation of proteins in intuitive conditions. To address these limitations,we propose CMADiff here, a novel framework that enables controllable protein generation by aligning the physicochemical properties of protein sequences with text-based descriptions through a latent diffusion process. Specifically, CMADiff employs a Conditional Variational Autoencoder (CVAE) to integrate physicochemical features as conditional input, forming a robust latent space that captures biological traits. In this latent space, we apply a conditional diffusion process, which is guided by BioAligner, a contrastive learning-based module that aligns text descriptions with protein features, enabling text-driven control over protein sequence generation. Validated by a series of evaluations including AlphaFold3, the experimental results indicate that CMADiff outperforms protein sequence generation benchmarks and holds strong potential for future applications. The implementation and code are available at https://github.com/HPC-NEAU/PhysChemDiff.

Problem

• Generates proteins with controlled physicochemical properties
• Aligns text descriptions with protein features for control
• Improves protein sequence generation using cross-modal diffusion

Innovation

• Aligns physicochemical properties with text descriptions
• Uses Conditional Variational Autoencoder for feature integration
• Employs BioAligner for text-driven protein generation control