UniSite: The First Cross-Structure Dataset and Learning Framework for End-to-End Ligand Binding Site Detection

2025-06-03

AI Summary

To address three key challenges in protein–ligand binding site detection—structural heterogeneity across complexes, fragmented processing pipelines, and inaccurate evaluation metrics—this paper introduces UniSite-DS, the first UniProt-centric multi-binding-site dataset, scaling 4.81× larger in multi-site instances and 2.08× larger in total instances than prior largest datasets. We propose an end-to-end binding site detection framework that unifies segmentation and clustering via set-based prediction modeling and bipartite matching using the Hungarian algorithm. Furthermore, we adopt IoU-weighted mean average precision (mAP) as a more robust evaluation metric. Experiments demonstrate that UniSite consistently outperforms state-of-the-art methods across multiple benchmarks, significantly mitigating statistical bias while improving detection consistency and evaluation reliability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity

Abstract

The detection of ligand binding sites for proteins is a fundamental step in Structure-Based Drug Design. Despite notable advances in recent years, existing methods, datasets, and evaluation metrics are confronted with several key challenges: (1) current datasets and methods are centered on individual protein-ligand complexes and neglect that diverse binding sites may exist across multiple complexes of the same protein, introducing significant statistical bias; (2) ligand binding site detection is typically modeled as a discontinuous workflow, employing binary segmentation and subsequent clustering algorithms; (3) traditional evaluation metrics do not adequately reflect the actual performance of different binding site prediction methods. To address these issues, we first introduce UniSite-DS, the first UniProt (Unique Protein)-centric ligand binding site dataset, which contains 4.81 times more multi-site data and 2.08 times more overall data compared to the previously most widely used datasets. We then propose UniSite, the first end-to-end ligand binding site detection framework supervised by set prediction loss with bijective matching. In addition, we introduce Average Precision based on Intersection over Union (IoU) as a more accurate evaluation metric for ligand binding site prediction. Extensive experiments on UniSite-DS and several representative benchmark datasets demonstrate that IoU-based Average Precision provides a more accurate reflection of prediction quality, and that UniSite outperforms current state-of-the-art methods in ligand binding site detection. The dataset and codes will be made publicly available at https://github.com/quanlin-wu/unisite.

Problem

• Detects diverse binding sites across multiple protein complexes
• Replaces discontinuous workflow with end-to-end detection framework
• Introduces improved evaluation metric for binding site prediction

Innovation

• First UniProt-centric ligand binding site dataset
• End-to-end detection framework with set prediction
• IoU-based Average Precision for accurate evaluation