Crowdsourced mapping of unexplored target space of kinase inhibitors
Anna Cichońska, Balaguru Ravikumar, Robert J. Allaway, Fang Wan, Sung‐Joon Park, Olexandr Isayev, Shuya Li, Mike J. Mason, Andrew Lamb, Ziaurrehman Tanoli, Minji Jeon, Sunkyu Kim, Mariya Popova, Stephen J. Capuzzi, Jianyang Zeng, Kristen K. Dang, Gregory Koytiger, Jaewoo Kang, Carrow I. Wells, Timothy M. Willson, User oselot, Mehmet Tan, Team N121, Chih-Han Huang, Edward S.C. Shih, Tsai‐Min Chen, Chih‐Hsun Wu, Wei-Quan Fang, Jhih-Yu Chen, Ming‐Jing Hwang, Team Let_Data_Talk, Xiaokang Wang, Marouen Ben Guebila, Behrouz Shamsaei, Sourav Singh, User thinng, Thin Nguyen, Team KKT, Mostafa Karimi, Di Wu, Zhangyang Wang, Yang Shen, Team Boun, Hakime Öztürk, Elif Özkırımlı, Arzucan Özgür, Team KinaseHunter, Hansaim Lim, Lei Xie, Team AmsterdamUMC-KU-team, Georgi K. Kanev, Albert J. Kooistra, Bart A. Westerman, Team DruginaseLearning, P.J. Terzopoulos, Konstantinos Ntagiantas, Christos Fotis, Leonidas G. Alexopoulos, Dimitri Boeckaerts, Michiel Stock, Bernard De Baets, Yves Briers, Team QED, Yunan Luo, Hailin Hu, Jian Peng, Team METU_EMBLEBI_CROssBAR, Tunca Doğan, Ahmet Süreyya Rifaioğlu, Heval Ataş, Rengül Çetin-Atalay, Volkan Atalay, María Martin, Team DMIS_DK, Minji Jeon, Junhyun Lee, Seongjun Yun, Bumsoo Kim, Buru Chang, Team AI Winter is Coming, Team hulab, Gábor Turu, Ádám Misák, Bence Szalai, László Hunyady, Team ML-Med, Matthias Lienhard, Paul Prasse, Ivo Bachmann, Julia Ganzlin, Gal Barel, Ralf Herwig, Team Prospectors, Davor Oršolić, Bono Lučić, Višnja Stepanić, Tomislav Šmuc, Challenge organizers, Tudor I. Oprea, Avner Schlessinger
Abstract
Despite decades of intensive search for compounds that modulate the activity of particular protein targets, a large proportion of the human kinome remains as yet undrugged. Effective approaches are therefore required to map the massive space of unexplored compound-kinase interactions for novel and potent activities. Here, we carry out a crowdsourced benchmarking of predictive algorithms for kinase inhibitor potencies across multiple kinase families tested on unpublished bioactivity data. We find the top-performing predictions are based on various models, including kernel learning, gradient boosting and deep learning, and their ensemble leads to a predictive accuracy exceeding that of single-dose kinase activity assays. We design experiments based on the model predictions and identify unexpected activities even for under-studied kinases, thereby accelerating experimental mapping efforts. The open-source prediction algorithms together with the bioactivities between 95 compounds and 295 kinases provide a resource for benchmarking prediction algorithms and for extending the druggable kinome.