MLScan-LCK: A machine-learning framework for screening of potential inhibitors targeting LCK protein

MLScan-LCK is a user-friendly, Tkinter-based desktop application for machine learning-driven screening of small molecules targeting the LCK protein, a key therapeutic target in anticancer drug discovery. The tool integrates six state-of-the-art classifiers such as Random Forest, Support Vector Machine, K-Nearest Neighbors, Extra Trees, LightGBM, and XGBoost, to predict compound bioactivity from SMILES input. To enhance prediction robustness and reliability, MLScan-LCK employs an ensemble soft-voting strategy, combining outputs from all six models into a single consensus prediction. The application supports both single-molecule evaluation and large-scale batch screening. In addition to bioactivity prediction, MLScan-LCK computes essential drug-likeness metrics, including Lipinski’s rule-of-five parameters and QED scores, enabling rapid prioritization of lead-like compounds. Benchmark testing using 1,000 natural products from the LOTUS database completed in 1 minute 49 seconds on a Ryzen 7 7700 system, with projected scalability to over 276,000 compounds within ~8.4 hours. MLScan-LCK is suitable for early-stage drug discovery, virtual screening, and in silico lead identification, and is distributed as a standalone Windows executable and a Linux-compatible Python package.