CFP-Gen: Combinatorial Functional Protein Generation via Diffusion Language Models

2025-05-28

AI Summary

Current protein language models (PLMs) struggle to jointly satisfy multimodal, multi-granularity constraints—including functional annotations (GO/EC/IPR), sequence features, and 3D structural properties. To address this, we propose the first diffusion-based language model framework enabling cross-modal joint guidance for de novo protein design. Our approach innovatively integrates annotation-guided functional modulation (AGFM) with residue-level functional encoding (RCFE), and incorporates a plug-and-play 3D structure encoder—enabling composable, controllable, and functionally multifaceted protein generation. Experiments demonstrate that generated proteins match natural counterparts in functional performance, with significantly improved success rates in multifunctional design. The framework achieves high throughput, high accuracy, and intrinsic interpretability. By unifying functional, sequential, and structural constraints within a single generative paradigm, our method establishes a new foundation for multi-constrained protein engineering.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Diffusion Models for Vision

Application Category

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

Abstract

Existing PLMs generate protein sequences based on a single-condition constraint from a specific modality, struggling to simultaneously satisfy multiple constraints across different modalities. In this work, we introduce CFP-Gen, a novel diffusion language model for Combinatorial Functional Protein GENeration. CFP-Gen facilitates the de novo protein design by integrating multimodal conditions with functional, sequence, and structural constraints. Specifically, an Annotation-Guided Feature Modulation (AGFM) module is introduced to dynamically adjust the protein feature distribution based on composable functional annotations, e.g., GO terms, IPR domains and EC numbers. Meanwhile, the Residue-Controlled Functional Encoding (RCFE) module captures residue-wise interaction to ensure more precise control. Additionally, off-the-shelf 3D structure encoders can be seamlessly integrated to impose geometric constraints. We demonstrate that CFP-Gen enables high-throughput generation of novel proteins with functionality comparable to natural proteins, while achieving a high success rate in designing multifunctional proteins. Code and data available at https://github.com/yinjunbo/cfpgen.

Problem

• Generates proteins with multiple cross-modal constraints
• Integrates functional, sequence, and structural constraints
• Enables high-throughput multifunctional protein design

Innovation

• Diffusion model integrates multimodal protein constraints
• AGFM module dynamically adjusts protein features
• RCFE module ensures precise residue control