Enhancing Multimodal Protein Function Prediction Through Dual-Branch Dynamic Selection with Reconstructive Pre-Training

2025-09-01

International Joint Conference on Artificial Intelligence

AI Summary

Modeling complex cross-modal dependencies among multimodal protein features—such as sequences, 3D structures, interaction networks, and physicochemical properties—remains challenging, limiting functional annotation accuracy. To address this, we propose a dual-branch dynamic selection and reconstruction-based pretraining framework. Methodologically: (1) reconstruction-based pretraining captures low-level semantic representations; (2) a bidirectional interaction module (BInM) enables deep cross-modal feature fusion; and (3) a dynamic selection module (DSM) adaptively tailors feature representations to diverse functional categories. Integrated with a hierarchical multi-label classification strategy, our framework achieves state-of-the-art performance on the human protein dataset across all three Gene Ontology (GO) domains—Biological Process (BPO), Molecular Function (MFO), and Cellular Component (CCO)—outperforming existing methods. It significantly enhances fine-grained functional annotation accuracy and model generalizability.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Multimodal protein features play a crucial role in protein function prediction. However, these features encompass a wide range of information, ranging from structural data and sequence features to protein attributes and interaction networks, making it challenging to decipher their complex interconnections. In this work, we propose a multimodal protein function prediction method (DSRPGO) by utilizing dynamic selection and reconstructive pre-training mechanisms. To acquire complex protein information, we introduce reconstructive pre-training to mine more fine-grained information with low semantic levels. Moreover, we put forward the Bidirectional Interaction Module (BInM) to facilitate interactive learning among multimodal features. Additionally, to address the difficulty of hierarchical multi-label classification in this task, a Dynamic Selection Module (DSM) is designed to select the feature representation that is most conducive to current protein function prediction. Our proposed DSRPGO model improves significantly in BPO, MFO, and CCO on human datasets, thereby outperforming other benchmark models.

Problem

• Decoding complex interconnections between diverse multimodal protein features
• Addressing hierarchical multi-label classification challenges in function prediction
• Enhancing feature selection for optimal protein function prediction performance

Innovation

• Reconstructive pre-training for fine-grained protein information
• Bidirectional Interaction Module for multimodal feature learning
• Dynamic Selection Module for optimal feature representation