![Author ORCID: We display the ORCID iD icon alongside authors names on our website to acknowledge that the ORCiD has been authenticated when entered by the user. To view the users ORCiD record click the icon. [opens in a new tab]](https://chemrxiv.org/engage/assets/public/chemrxiv/images/logos/orcid.png)
Abstract
Kinase inhibitors are crucial in cancer treatment, but drug resistance and side effects hinder the development of effective drugs. To address these challenges, it is essential to analyze the polypharmacology of kinase inhibitor and identify compound with high selectivity profile. This study presents KinomeMETA, a framework for profiling the activity of small molecule kinase inhibitors across a panel of 661 kinases. By training a meta-learner based on a graph neural network and fine-tuning it to create kinase-specific learners, KinomeMETA outperforms benchmark multi-task models and other kinase profiling models. It provides higher accuracy for understudied kinases with limited known data and broader coverage of kinase types, including important mutant kinases. Case studies on the discovery of new scaffold inhibitors for PKMYT1 and selective inhibitors for drug-resistant mutants of FGFRs demonstrate the role of KinomeMETA in virtual screening and kinome-wide activity profiling. Overall, KinomeMETA has the potential to accelerate kinase drug discovery by more effectively exploring the kinase polypharmacology landscape.
Supplementary materials
Title
Supplementary information
Description
Fig. S1 | The distribution of datasets before and after negative sampling. Fig. S2 | Performance of meta-learner build in the cluster-wise manner. Fig. S3 | Model performance for different hyperparameters. Fig. S4 | Labels of samples in KIT and its mutant forms. Fig. S5 | The performance of KinomeMETA for ABL1, ALK, RET, MET, EGFR and FLT3 mutants compared with MTGNN and “SameAsWild”. all measured by MCC. Fig. S6 | Performances comparison between KinomeMETA and MTGNN in the data-adding experiment, in terms of auROC. Fig. S7 | Training strategy of MTGNN. Fig. S8 | Dose response curves used to generate IC50 for PKMYT1 inhibitors. Table S1. The Hyperparameter settings for KinomeMETA. Table S2. The parameter settings for benchmark models. Table S3. Comparing KinomeMETA’s performance with RF, MT-DNN, IDDkin, ST-DNN and Auto-Sklearn. Table S4. Statistics of data and model performance of FGFRs’ models.
Actions
Title
supplementary tabel 5, 6 and 7
Description
Table S5.Kinase information. Table S6. PKMYT1 screening results. Table S7. Experimental and prediction results for compound 2e and compound 15.
Actions
