KiSSim: Predicting Off-Targets from Structural Similarities in the Kinome

Area of Science:

• Biochemistry and structural biology
• Pharmacology and drug discovery

Background:

• Protein kinases are crucial drug targets, but developing selective inhibitors is difficult due to conserved binding sites.
• Dysregulated kinases are implicated in cancer, inflammatory, and degenerative diseases.

Purpose of the Study:

• To develop a novel computational method for predicting kinase off-targets.
• To enhance the understanding of kinase structural similarities and their implications for drug selectivity.

Main Methods:

• Introduction of the kinase-focused, subpocket-enhanced KiSSim fingerprint (Kinase Structural Similarity).
• Utilizing the KLIFS pocket definition for residue-by-residue comparison across protein kinase structures.
• Encoding physicochemical and spatial properties within structural context, including key subpockets.

Main Results:

• KiSSim identified unexpected off-targets by analyzing structural similarities, outperforming sequence-based classifications.
• Calculated all-against-all similarities within the kinome, revealing high structural similarity between EGFR and off-targets SLK and LOK.
• KiSSim demonstrated comparable or superior performance to existing methods like KLIFS sequence identity, IFPs, and SiteAlign.

Conclusions:

• The KiSSim fingerprint is a valuable tool for predicting kinase off-targets and polypharmacology.
• Structural information provides complementary insights to sequence data for kinase inhibitor design.
• The method is available as an open-source Python package and conda package.