KiSSim

KiSSim computes kinase structural similarity fingerprints to predict off-target interactions among protein kinases and support polypharmacology analyses.

Key Features:

• Kinase-focused, subpocket-enhanced fingerprint: The KiSSim fingerprint encodes physicochemical and spatial properties of kinase pocket residues to capture structural context.
• KLIFS pocket alignment: Uses the KLIFS pocket definition of 85 residues aligned across available protein kinase structures to enable residue-by-residue comparison without structural alignment.
• Subpocket emphasis: Specifically encodes features of key subpockets including the hinge region, the DFG motif, and the front pocket.
• Structure and sequence integration: Combines structural and sequence information to identify off-targets not apparent from sequence-based kinome tree groupings.
• All-against-all similarity computation: Calculates all-against-all similarities within the structurally covered kinome for off-target prediction.
• Comparative benchmarking: Similarity outputs have been compared to KLIFS pocket sequence identity, KLIFS interaction fingerprints (IFPs), and SiteAlign.
• 3D mapping of fingerprints: Fingerprint values can be visualized on structures by coloring at residue and feature resolution to aid interpretation.

Scientific Applications:

• Off-target prediction: Predicts off-target effects among protein kinases based on pocket structural similarity.
• Polypharmacology analysis: Guides polypharmacology predictions by identifying structurally similar kinase targets across the kinome.
• Structural rationalization: Rationalizes observed structural similarities and differences by mapping fingerprint features onto 3D kinase structures.
• Case example: Identified unexpected off-targets for Erlobinib such as SLK and LOK that show high structural similarity to EGFR despite belonging to a different kinome group.

Methodology:

Implements a kinase-focused, subpocket-enhanced fingerprint using the KLIFS 85-residue pocket alignment, encodes physicochemical and spatial residue features with subpocket focus (hinge, DFG, front pocket), performs residue-by-residue comparisons without structural alignment, and computes all-against-all kinome similarities compared against KLIFS pocket sequence identity, KLIFS IFPs, and SiteAlign.