KEPLA: A Knowledge-Enhanced Deep Learning Framework for Accurate Protein-Ligand Binding Affinity Prediction

2025-06-16

AI Summary

Accurately predicting protein–ligand binding affinity remains a critical challenge in drug discovery. Existing deep learning methods over-rely on 3D structural features while neglecting prior biological knowledge—such as Gene Ontology (GO) annotations and ligand biochemical properties—leading to limited generalizability and interpretability. To address this, we propose KALP, the first knowledge graph (KG)-enhanced model that jointly integrates GO annotations and ligand physicochemical knowledge. KALP innovatively introduces a joint optimization framework combining KG relation alignment and local cross-attention, enabling global semantic alignment between protein sequences and ligand molecular graphs, as well as fine-grained cross-modal representation learning. The model adopts a dual-objective multi-task architecture that unifies KG embedding learning with cross-modal attention. On both in-domain and cross-domain benchmarks, KALP significantly outperforms state-of-the-art methods. Visualization and ablation studies further validate its biological interpretability and knowledge-aware reasoning capability.

Technology Category

• Knowledge Representation and Reasoning: Description Logics
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Machine Learning: Graph-based Machine Learning

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: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Accurate prediction of protein-ligand binding affinity is critical for drug discovery. While recent deep learning approaches have demonstrated promising results, they often rely solely on structural features, overlooking valuable biochemical knowledge associated with binding affinity. To address this limitation, we propose KEPLA, a novel deep learning framework that explicitly integrates prior knowledge from Gene Ontology and ligand properties of proteins and ligands to enhance prediction performance. KEPLA takes protein sequences and ligand molecular graphs as input and optimizes two complementary objectives: (1) aligning global representations with knowledge graph relations to capture domain-specific biochemical insights, and (2) leveraging cross attention between local representations to construct fine-grained joint embeddings for prediction. Experiments on two benchmark datasets across both in-domain and cross-domain scenarios demonstrate that KEPLA consistently outperforms state-of-the-art baselines. Furthermore, interpretability analyses based on knowledge graph relations and cross attention maps provide valuable insights into the underlying predictive mechanisms.

Problem

• Enhances protein-ligand binding affinity prediction accuracy
• Integrates Gene Ontology and ligand properties knowledge
• Improves interpretability and cross-domain prediction performance

Innovation

• Integrates Gene Ontology and ligand properties
• Aligns global representations with knowledge graphs
• Uses cross attention for fine-grained embeddings