RT Journal Article
SR Electronic
T1 Crowdsourced mapping of unexplored target space of kinase inhibitors
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2019.12.31.891812
DO 10.1101/2019.12.31.891812
A1 Anna Cichonska
A1 Balaguru Ravikumar
A1 Robert J Allaway
A1 Sungjoon Park
A1 Fangping Wan
A1 Olexandr Isayev
A1 Shuya Li
A1 Michael Mason
A1 Andrew Lamb
A1 Zia-ur-Rehman Tanoli
A1 Minji Jeon
A1 Sunkyu Kim
A1 Mariya Popova
A1 Jianyang Zeng
A1 Kristen Dang
A1 Gregory Koytiger
A1 Jaewoo Kang
A1 Carrow I. Wells
A1 Timothy M. Willson
A1 The IDG-DREAM Drug-Kinase Binding Prediction Challenge Consortium
A1 Tudor I. Oprea
A1 Avner Schlessinger
A1 David H. Drewry
A1 Gustavo Stolovitzky
A1 Krister Wennerberg
A1 Justin Guinney
A1 Tero Aittokallio
YR 2020
UL http://biorxiv.org/content/early/2020/01/07/2019.12.31.891812.abstract
AB Despite decades of intensive search for compounds that modulate the activity of particular proteins, there are currently small-molecule probes available only for a small proportion of the human proteome. Effective approaches are therefore required to map the massive space of unexplored compound-target interactions for novel and potent activities. Here, we carried out a crowdsourced benchmarking of the accuracy of machine learning (ML) algorithms at predicting kinase inhibitor potencies across multiple kinase families. A total of 268 ML predictions were scored in unpublished bioactivity data sets. Top-performing algorithms used kernel learning, gradient boosting and deep learning, with predictive accuracy exceeding that of target activity assays. Subsequent experiments carried out based on the the top-performing model predictions demonstrated that these models and their ensemble can improve the accuracy of experimental mapping efforts, especially for so far under-studied kinases. The open-source ML algorithms together with the novel dose-response data for 905 bioactivities between 95 compounds and 295 kinases provide a unique resource for extending the druggable kinome.