LCK-SafeScreen-Model: An Advanced Ensemble Machine Learning Approach for Estimating the Binding Affinity between...

Area of Science:

• Biochemistry
• Computational Chemistry
• Pharmacology

Background:

• Lymphocyte-specific protein tyrosine kinase (LCK) is a key target in leukemia therapy.
• Off-target interactions of LCK inhibitors can cause adverse effects, necessitating accurate prediction methods.
• Reliable prediction of drug-LCK interactions is crucial during drug development.

Purpose of the Study:

• To develop and optimize an ensemble machine learning model for predicting binding affinity between drug molecules and LCK.
• To enhance the accuracy of LCK inhibitor prediction for drug discovery and safety assessment.
• To provide accessible tools (webserver, GitHub) for researchers.

Main Methods:

• Generation and selection of molecular fingerprints.
• Design and hyperparameter tuning of an ensemble machine learning model.
• Screening of a drug library, experimental validation using ADP-Glo assay, and molecular docking.

Main Results:

• The ensemble model achieved improved predictive performance, increasing test R-squared from 0.644 to 0.730.
• Test Root Mean Squared Error (RMSE) was reduced from 0.841 to 0.732.
• The model successfully identified top-scoring compounds from a drug library, validated by experimental assays and docking.

Conclusions:

• The refined ensemble model demonstrates high accuracy in predicting LCK inhibitors.
• This approach is effective for both safety panel predictions and the discovery of novel LCK inhibitors.
• The developed tools facilitate broader application and research in LCK-targeted therapies.