DeepUMQA3: a web server for accurate assessment of interface residue accuracy in protein complexes

• DOI: 10.1093/bioinformatics/btad591
• Publication Date: 2023-09-23T04:04:50Z
• Authors: Jun Liu; Dong Liu; Gui-Jun Zhang
• Keywords: Interface (matter); Residue (chemistry)

Model quality assessment is a crucial part of protein structure prediction and gateway to proper usage models in biomedical applications. Many methods have been proposed for assessing the structural monomers, but few evaluating complex models. As becomes new challenge, there an urgent need model that can accurately assess accuracy interface residues structures.Here, we present DeepUMQA3, web server structures using deep neural networks. For input structure, features are extracted from three levels overall complex, intra-monomer, inter-monomer, improved residual network used predict per-residue lDDT residue accuracy. DeepUMQA3 ranks first blind test estimation CASP15, with Pearson, Spearman, AUC 0.564, 0.535, 0.755 under measurement, which 17.6%, 23.6%, 10.9% higher than second best method, respectively. also all entire distinguish high- low-precision residues.The sever freely available at http://zhanglab-bioinf.com/DeepUMQA_server/.