Impact of AlphaFold on structure prediction of protein complexes: The CASP15-CAPRI experiment

We present the results for CAPRI Round 54, the 5th joint CASP-CAPRI protein assembly prediction challenge. The Round offered 37 targets, including 14 homodimers, 3 homo-trimers, 13 heterodimers including 3 antibody-antigen complexes, and 7 large assemblies. On average ~70 CASP and CAPRI predictor groups, including more than 20 automatics servers, submitted models for each target. A total of 21 941 models submitted by these groups and by 15 CAPRI scorer groups were evaluated using the CAPRI model quality measures and the DockQ score consolidating these measures. The prediction performance was quantified by a weighted score based on the number of models of acceptable quality or higher submitted by each group among their five best models. Results show substantial progress achieved across a significant fraction of the 60+ participating groups. High-quality models were produced for about 40% of the targets compared to 8% two years earlier. This remarkable improvement is due to the wide use of the AlphaFold2 and AlphaFold2-Multimer software and the confidence metrics they provide. Notably, expanded sampling of candidate solutions by manipulating these deep learning inference engines, enriching multiple sequence alignments, or integration of advanced modeling tools, enabled top performing groups to exceed the performance of a standard AlphaFold2-Multimer version used as a yard stick. This notwithstanding, performance remained poor for complexes with antibodies and nanobodies, where evolutionary relationships between the binding partners are lacking, and for complexes featuring conformational flexibility, clearly indicating that the prediction of protein complexes remains a challenging problem.

Paul A. Bates: This work was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC0001003), the UK Medical Research Council (FC001003), and the Wellcome Trust (FC001003). Alexandre M. J. J. Bonvin: Financial support from the European Union Horizon 2020, project BioExcel (823830) and from the Netherlands e-Science Center, project IVRESSE (027.020.G13) is acknowledged. Jianlin Cheng: This work was partially supported by US National Science Foundation (grant #: DBI1759934 and IIS1763246), and US National Institutes of Health (grant #: R01GM093123 and R01GM146340). Juan Fernandez-Recio: Spanish Ministry of Science (grant PID2019-110167RB-I00/AEI/10.13039/501100011033). Jeffrey J. Gray: This work was supported by National Institute of Health grant R35-GM141881. Computational resources were provided by the Advanced Research Computing at Hopkins (ARCH) core facility, supported by the National Science Foundation (NSF) grant number OAC1920103. Sheng-You Huang: This work was supported by the National Natural Science Foundation of China to Sheng-You Huang (Grant No: 62072199 and 32161133002). Brian Jimenez-Garcia: Jorge Roel-Touris acknowledges funding from the European Molecular Biology Organization (EMBO) [ALTF 145-2021]; Brian Jimenez-Garcia is employed by Zymvol Biomodeling on a project which received funding from the European Union's Horizon 2020 research and innovation programme under Marie Skłodowska-Curie grant agreement No. 801342 (Tecniospring INDUSTRY) and the Government of Catalonia's Agency for Business Competitiveness (ACCIO). Daisuke Kihara: We are grateful for ITaP Research Computing at Purdue University for providing us additional computational resources for this project. This work is partially supported by the National Institutes of Health (R01GM133840, R01GM123055), the National Science Foundation (CMMI1825941, MCB1925643, DBI2146026, IIS2211598, and DBI2003635). Cezary Czaplewski and Jacob C. Verburgt were supported by a National Institute of General Medical Sciences-funded predoctoral fellowship to Cezary Czaplewski and Jacob C. Verburgt (T32 GM132024). Kozakov/Vajda: DMS 2054251 from the National Science Foundation, and R01GM140098, R35GM118078, RM1135136 from the National Institute of Health. Romina Oliva: This research was funded by the KAUST baseline research funding (to Luigi Cavallo) and by MIUR-FFABR “Fondoper il Finanziamento Attivita Base di Ricerca” (to Romina Oliva). Brian G. Pierce: This work was supported by NIH R35 GM144083 (to Brian G. Pierce). Yang Shen: This study was supported by NIH/NIGMS (R35GM124952 to Yang Shen) and NSF (CCF-1943008 to Yang Shen). Portions of this research were conducted with the advanced computing resources provided by Texas A&M High Performance Research Computing. Adam K. Sieradzan: Supported by National Science Center of Poland (Narodowe Centrum Nauki) (NCN), grants UMO2017/25/B/ST4/01026, UMO-2017/26/M/ST4/00044, and UMO-2017/27/B/ST4/00926. Computational resources were provided by (a) the Interdisciplinary Centre of Mathematical and Computational Modelling (ICM) at the University of Warsaw (b) the Centre of Informatics—Tricity Academic Superkomputer & NetworK (CI TASK) in Gdansk, (c) the Polish Grid Infrastructure (PL-GRID), and (d) our Beowulf cluster at the Faculty of Chemistry, University of Gdansk. Ceslovas Venclovas: Research Council of Lithuania (grant S-MIP-21-25). Bjorn Wallner: Wallenberg AI, Autonomous System and Software Program (WASP) and Berzelius from Knut and Alice Wallenberg Foundation (KAW), Swedish Research Council, 2020-03352, The Swedish e-Science Research Center, and Carl Tryggers stiftelse for Vetenskaplig Forskning, 20:453. Lin Yang: The authors acknowledge the financial support from the Science Foundation of the National Key Laboratory of Science and Technology on Advanced Composites in Special Environments, the Fundamental Research Funds for the Central Universities of China. Xiaoqin Zou: This was supported by NIH/NIGMS R35GM136409 to Xiaoqin Zou.

Peer reviewed