TY - JOUR

T1 - Predicting the target landscape of kinase inhibitors using 3D convolutional neural networks

AU - Kanev, Georgi K.

AU - Zhang, Yaran

AU - Kooistra, Albert J.

AU - Bender, Andreas

AU - Leurs, Rob

AU - Bailey, David

AU - Würdinger, Thomas

AU - Graaf, Chris de

AU - de Esch, Iwan J.P.

AU - Westerman, Bart A.

N1 - Funding Information: Funding: G.K., B.A.W. C. d. G., T.W., R.L. and I.d.E. are supported by Amsterdam Data Science (ADS), B.A.W., D.B. and T.W. are supported by the Brain Tumour Charity Grant 488097 (WINDOW consortium), B.A.W. is supported by the Innovation Exchange Amsterdam (IXA) grant APCA-PoC-2017 and G.K. is supported by the Cancer Center Amsterdam (GlioSPARK project 2018-2-19). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. None of the funders played any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Publisher Copyright: Copyright: © 2023 Kanev et al.

PY - 2023

Y1 - 2023

N2 - Many therapies in clinical trials are based on single drug-single target relationships. To further extend this concept to multi-target approaches using multi-targeted drugs, we developed a machine learning pipeline to unravel the target landscape of kinase inhibitors. This pipeline, which we call 3D-KINEssence, uses a new type of protein fingerprints (3D FP) based on the structure of kinases generated through a 3D convolutional neural network (3D-CNN). These 3D-CNN kinase fingerprints were matched to molecular Morgan fingerprints to predict the targets of each respective kinase inhibitor based on available bioactivity data. The performance of the pipeline was evaluated on two test sets: a sparse drug-target set where each drug is matched in most cases to a single target and also on a densely-covered drug-target set where each drug is matched to most if not all targets. This latter set is more challenging to train, given its non-exclusive character. Our model’s root-mean-square error (RMSE) based on the two datasets was 0.68 and 0.8, respectively. These results indicate that 3D FP can predict the target landscape of kinase inhibitors at around 0.8 log units of bioactivity. Our strategy can be utilized in proteochemometric or chemogenomic workflows by consolidating the target landscape of kinase inhibitors.

AB - Many therapies in clinical trials are based on single drug-single target relationships. To further extend this concept to multi-target approaches using multi-targeted drugs, we developed a machine learning pipeline to unravel the target landscape of kinase inhibitors. This pipeline, which we call 3D-KINEssence, uses a new type of protein fingerprints (3D FP) based on the structure of kinases generated through a 3D convolutional neural network (3D-CNN). These 3D-CNN kinase fingerprints were matched to molecular Morgan fingerprints to predict the targets of each respective kinase inhibitor based on available bioactivity data. The performance of the pipeline was evaluated on two test sets: a sparse drug-target set where each drug is matched in most cases to a single target and also on a densely-covered drug-target set where each drug is matched to most if not all targets. This latter set is more challenging to train, given its non-exclusive character. Our model’s root-mean-square error (RMSE) based on the two datasets was 0.68 and 0.8, respectively. These results indicate that 3D FP can predict the target landscape of kinase inhibitors at around 0.8 log units of bioactivity. Our strategy can be utilized in proteochemometric or chemogenomic workflows by consolidating the target landscape of kinase inhibitors.

U2 - 10.1371/journal.pcbi.1011301

DO - 10.1371/journal.pcbi.1011301

M3 - Journal article

C2 - 37669273

AN - SCOPUS:85171807183

VL - 19

JO - P L o S Computational Biology (Online)

JF - P L o S Computational Biology (Online)

SN - 1553-734X

IS - 9

M1 - e1011301

ER -