MetaScore - 机器学习评分改进方法 - 用于蛋白-蛋白对接构象评分

Aug 5, 2026

Upload by Jiven

Protein–protein interactions are central to biological function, and determining the 3D structures of protein complexes is crucial for understanding their structural basis and cellular coordination. Experimental structure determination is expensive and slow, so computational docking is widely used, but it still faces the scoring challenge: selecting near-native models from many docked conformations. MetaScore introduces a random-forest based machine-learning scorer using protein–protein interfacial features, averaging RF predictions with traditional scoring functions. Results show consistent gains over nine traditional scoring functions and improved performance via a MetaScore-Ensemble strategy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages21

Reading Time53 min

AccessFree to download

Introduction

Protein–protein interactions and complex structures

Computational docking and the scoring problem

MetaScore method

Random-forest classifier and interfacial features

Integration with traditional scoring functions

Results and evaluation

Performance vs. traditional scoring functions

MetaScore-Ensemble comparison

Conclusion

Article MetaScore: A Novel MachineLearning-Based Approach to Improve Traditional Scoring Functions for Scoring Protein– Protein Docking Conformations Yong Jung, Cunliang Geng, Alexandre M. J. J. Bonvin, Li C. Xue and Vasant G. Honavar Special Issue Protein Structure Prediction with AlphaFold Edited by Dr. Philippe Urban and Dr. Denis Pompon [https://doi.org/10.3390/biom13010121](https://doi.org/10.3390/biom13010121) biomolecules Article MetaScore: A Novel Machine-Learning-Based Approach to Improve Traditional Scoring Functions for Scoring Protein–Protein Docking Conformations Yong Jung 1,2,3, Cunliang Geng 4, Alexandre M. J. J. Bonvin 4, Li C. Xue 4,5, * and Vasant G. Honavar 1,2,3,6,7,8,9, * Citation: Jung, Y.; Geng, C.; Bonvin, A.M.J.J.; Xue, L.C.; Honavar, V.G. MetaScore: A Novel Machine-Learning-Based Approach to Improve Traditional Scoring Functions for Scoring Protein–Protein Docking Conformations. Biomolecules 2023, 13, 121. [https://doi.org/](https://doi.org/)[ ](https://doi.org/)[10.3390/biom13010121](10.3390/biom13010121) Academic Editors: Philippe Urban and Denis Pompon Received: 1 December 2022 Revised: 22 December 2022 Accepted: 26 December 2022 Published: 6 January 2023 Copyright: © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 Bioinformatics & Genomics Graduate Program, Pennsylvania State University, University Park, PA 16802, USA 2 Arti􀀂cial Intelligence Research Laboratory, Pennsylvania State University, University Park, PA 16802, USA 3 Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA 4 Bijvoet Centre for Biomolecular Research, Faculty of Science—Chemistry, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands 5 Center for Molecular and Biomolecular Informatics, Radboudumc, Greet Grooteplein 26-28, 6525 GA Nijmegen, The Netherlands 6 Clinical and Translational Sciences Institute, Pennsylvania State University, University Park, PA 16802, USA 7 College of Information Sciences & Technology, Pennsylvania State University, University Park, PA 16802, USA 8 Institute for Computational and Data Sciences, Pennsylvania State University, University Park, PA 16802, USA 9 Center for Big Data Analytics and Discovery Informatics, Pennsylvania State University, University Park, PA 16823, USA * Correspondence: [li.xue@radboudumc.nl](li.xue@radboudumc.nl) (L.C.X.); [vhonavar@psu.edu](vhonavar@psu.edu) (V.G.H.); Tel.: +31-61-859-4390 (L.C.X.); +1-814-865-3141 (V.G.H.) Abstract: Protein–protein interactions play a ubiquitous role in biological function. Knowledge of the three-dimensional (3D) structures of the complexes they form is essential for understanding the structural basis of those interactions and how they orchestrate key cellular processes. Computational docking has become an indispensable alternative to the expensive and time-consuming experimental approaches for determining the 3D structures of protein complexes. Despite recent progress, identifying near-native models from a large set of conformations sampled by docking—the so-called scoring problem—still has considerable room for improvement. We present MetaScore, anew machine-learning-based approach to improve the scoring of docked conformations. MetaScore utilizes a random forest (RF) classi􀀂er trained to distinguish near-native from non-native conformations using their protein–protein interfacial features. The features include physicochemical properties, energy terms, interaction-propensity-based features, geometric properties, interface topology features, evolutionary conservation, and also scores produced by traditional scoring functions (SFs) .

FAQs

MetaScore解决的核心问题是什么？

MetaScore针对蛋白-蛋白对接中的“评分问题”：在大量对接构象中识别接近天然结构（near-native）的模型仍有提升空间。

MetaScore如何使用机器学习进行评分？

MetaScore使用随机森林（RF）分类器，根据蛋白-蛋白界面特征区分near-native与non-native构象。

MetaScore与传统评分函数如何结合？

MetaScore将RF分类器产生的分数与任意传统评分函数（SF）产生的分数进行平均，从而得到最终评分。

MetaScore-Ensemble的作用是什么？

MetaScore-Ensemble通过将MetaScore的多个变体（结合RF与不同传统SF）进行组合，在top 10排名的构象评估中优于各个MetaScore变体以及传统SF。

Aug 5, 2026

Upload by Jiven