Impact of <scp>AlphaFold</scp> on structure prediction of protein complexes: The <scp>CASP15‐CAPRI</scp> experiment
Marc F. Lensink, Guillaume Brysbaert, Nessim Raouraoua, Paul A. Bates, Marco Giulini, Rodrigo V. Honorato, Charlotte van Noort, João M. C. Teixeira, Alexandre M. J. J. Bonvin, Ren Kong, Hang Shi, Xufeng Lu, Shan Chang, Jian Liu, Zhiye Guo, Xiao Chen, Alex Morehead, Raj S. Roy, Tianqi Wu, Nabin Giri, Farhan Quadir, Chen Chen, Jianlin Cheng, Carlos A. Del Carpio, Eichiro Ichiishi, Luis Angel Rodríguez‐Lumbreras, Juan Fernández‐Recio, Ameya Harmalkar, Lee‐Shin Chu, Samuel W. Canner, Rituparna Smanta, Jeffrey J. Gray, Hao Li, Peicong Lin, Jiahua He, Huanyu Tao, Sheng‐You Huang, Jorge Roel‐Touris, Brian Jiménez‐García, Charles Christoffer, Anika Jain, Yuki Kagaya, Harini Kannan, Tsukasa Nakamura, Genki Terashi, Jacob Verburgt, Yuanyuan Zhang, Zicong Zhang, Hayato Fujuta, Masakazu Sekijima, Daisuke Kihara, Omeir Khan, Sergei Kotelnikov, Usman Ghani, Dzmitry Padhorny, Dmitri Beglov, Sándor Vajda, Dima Kozakov, Surendra S. Negi, Tiziana Ricciardelli, Didier Barradas‐Bautista, Zhen Cao, Mohit Chawla, Luigi Cavallo, Romina Oliva, Rui Yin, Melyssa Cheung, Johnathan D. Guest, Jessica Lee, Brian G. Pierce, Ben Shor, Tomer Cohen, Matan Halfon, Dina Schneidman‐Duhovny, Shaowen Zhu, Rujie Yin, Yuanfei Sun, Yang Shen, Martyna Maszota‐Zieleniak, Krzysztof K. Bojarski, Emilia A. Lubecka, Mateusz Marcisz, Annemarie Danielsson, Łukasz Dziadek, Margrethe Gaardløs, Artur Giełdoń, Adam Liwo, Sergey A. Samsonov, Rafał Ślusarz, Karolina Zięba, Adam K. Sieradzan, Cezary Czaplewski, Shinpei Kobayashi, Yuta Miyakawa, Yasuomi Kiyota, Mayuko Takeda‐Shitaka, Kliment Olechnovič, Lukas Valančauskas, Justas Dapkūnas, Česlovas Venclovas
Abstract
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.