SUSPECT-ABL

SUSPECT-ABL predicts drug resistance mutations in the Abelson 1 (ABL1) kinase and quantifies mutation-induced changes in inhibitor binding free energy (ΔΔG) to inform resistance mechanisms.

Key Features:

• Structure-guided machine learning: Uses structure-guided machine learning to predict ABL1 mutations that confer drug resistance.
• ΔΔG estimation: Calculates changes in ligand binding free energy (ΔΔG) for ABL1 mutations.
• Drug-specific assessment: Assesses mutation effects on binding affinity for eight FDA-approved ABL1 inhibitors and distinguishes impacts on type I versus type II inhibitors.
• In silico saturation mutagenesis: Performs in silico saturation mutagenesis across ABL1 to evaluate all possible single-residue substitutions.
• Performance and validation: Validated on non-redundant blind tests with Matthew's Correlation Coefficient up to 0.73 for resistance classification and Pearson correlation up to 0.77 for ΔΔG prediction.
• Resistance profiling and prioritization: Generates resistance profiles and prioritizes potential emerging resistance mutations for further experimental validation and inhibitor design.

Scientific Applications:

• Clinical variant interpretation: Inform interpretation of ABL1 variants in chronic myeloid leukemia (CML) and other ABL1-related contexts.
• Precision oncology: Support precision medicine decisions by predicting likely resistance to approved ABL1 inhibitors.
• Drug discovery: Guide rational design of next-generation ABL1 inhibitors to mitigate resistance.
• Experimental prioritization: Prioritize candidate mutations for in vivo experimental validation based on predicted ΔΔG and resistance potential.

Methodology:

Employs structure-guided machine learning models to predict resistance, computes mutation-induced ΔΔG changes, performs in silico saturation mutagenesis across ABL1, differentiates effects on type I and type II inhibitors, and was evaluated using non-redundant blind tests.