A Multimodal Human Protein Embeddings Database: DeepDrug Protein Embeddings Bank (DPEB)

2025-10-24

AI Summary

Existing protein representation methods lack multimodal integration, limiting performance on downstream tasks such as protein–protein interaction (PPI) prediction. To address this, we construct the first unified embedding database for 22,043 human proteins, integrating four complementary modalities: AlphaFold2 internal-layer structural embeddings, BioEmbeddings sequence embeddings, ESM-2 contextual representations, and ProtVec n-gram statistical features—marking the first systematic utilization of AlphaFold2’s intermediate-layer embeddings. Our multimodal ensemble framework leverages multiple graph neural networks to fuse these representations. On PPI prediction, it achieves 87.37% AUROC and 79.16% accuracy; for enzyme classification and protein family classification, accuracies reach 77.42% and 86.04%, respectively. This database fills a critical gap in deep, transferable protein embedding resources, enabling scalable systems biology analysis and accelerating drug target discovery.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• 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

Abstract

Computationally predicting protein-protein interactions (PPIs) is challenging due to the lack of integrated, multimodal protein representations. DPEB is a curated collection of 22,043 human proteins that integrates four embedding types: structural (AlphaFold2), transformer-based sequence (BioEmbeddings), contextual amino acid patterns (ESM-2: Evolutionary Scale Modeling), and sequence-based n-gram statistics (ProtVec]). AlphaFold2 protein structures are available through public databases (e.g., AlphaFold2 Protein Structure Database), but the internal neural network embeddings are not. DPEB addresses this gap by providing AlphaFold2-derived embeddings for computational modeling. Our benchmark evaluations show GraphSAGE with BioEmbedding achieved the highest PPI prediction performance (87.37% AUROC, 79.16% accuracy). The framework also achieved 77.42% accuracy for enzyme classification and 86.04% accuracy for protein family classification. DPEB supports multiple graph neural network methods for PPI prediction, enabling applications in systems biology, drug target identification, pathway analysis, and disease mechanism studies.

Problem

• Lack of integrated multimodal protein representations for PPI prediction
• Missing AlphaFold2 neural network embeddings in public databases
• Need unified protein embeddings for computational biology applications

Innovation

• Integrates four multimodal protein embedding types
• Provides AlphaFold2-derived embeddings for computational modeling
• Supports graph neural networks for protein interaction prediction