TMLRpred: A machine learning classification model to distinguish reversible EGFR double mutant inhibitors

Area of Science:

• Oncology
• Computational Chemistry
• Drug Discovery

Background:

• Epidermal Growth Factor Receptor (EGFR) is a key target in lung cancer therapy.
• First-generation EGFR inhibitors are effective against L858R mutations but fail due to the T790M resistance mutation.
• There is a critical need for novel inhibitors targeting EGFR double mutants resistant to existing therapies.

Purpose of the Study:

• To develop and validate machine learning models for predicting potent EGFR inhibitors against resistant double mutants.
• To identify key structural features of active inhibitors through comparative analysis.
• To create an open-source tool for predicting potential anti-EGFR drug candidates.

Main Methods:

• Employed machine learning techniques including instance-based learner (IBK), naïve Bayesian (NB), sequential minimal optimization (SMO), and random forest (RF).
• Developed twelve classification models using three different datasets (high, moderate, and weakly active inhibitors).
• Validated models using fivefold cross-validation and independent datasets; analyzed functional groups, PubChem fingerprints, and substructures.

Main Results:

• Random forest-based models demonstrated the best performance in predicting EGFR inhibitor activity.
• Identified specific functional groups and substructures crucial for inhibitor activity against resistant mutants.
• Developed a functional tool incorporating the best models for predicting anti-TMLR inhibitor potential.

Conclusions:

• Machine learning classification models, particularly random forest, are effective in identifying novel inhibitors for resistant EGFR double mutants.
• The developed models and tool can accelerate the discovery of new anti-lung cancer drugs.
• This study contributes to open-source drug discovery efforts for overcoming therapeutic resistance.