Decoding kinase-adverse event associations for small molecule kinase inhibitors

Area of Science:

• Oncology
• Pharmacology
• Computational Biology

Background:

• Small molecule kinase inhibitors (SMKIs) are crucial in cancer treatment, often approved rapidly.
• Understanding the link between kinase targets and adverse events (AEs) is vital for patient safety.

Purpose of the Study:

• To develop a machine learning model to predict adverse events (AEs) associated with small molecule kinase inhibitors (SMKIs).
• To create a publicly accessible tool for identifying kinase-target adverse event pairs.
• To enhance precision medicine and drug development by assessing SMKI safety profiles.

Main Methods:

• Constructed a multi-domain dataset from 4638 patients across 16 FDA-approved SMKI registrational trials.
• Employed a machine learning model to analyze relationships between 442 kinases and 2145 AEs.
• Validated the model internally and externally using independent SMKI datasets.

Main Results:

• Systematically evaluated potential associations between 442 kinases and 2145 AEs.
• Developed an interactive web application, "Identification of Kinase-Specific Signal" (https://gongj.shinyapps.io/ml4ki).
• The model demonstrated utility in identifying and verifying kinase-adverse event pairs.

Conclusions:

• The developed model serves as a platform for discovering novel kinase-adverse event (KI-AE) pairs.
• It functions as a precision medicine tool to forecast clinical safety signals and mitigate patient risks.
• The tool supports modern drug development by enabling safety-focused screening and comparison of SMKI therapies.