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

The lymphocyte-specific protein tyrosine kinase (LCK) is a critical target in leukemia treatment. However, potential off-target interactions involving LCK can lead to unintended consequences. This underscores the importance of accurately predicting the inhibitory reactions of drug molecules with LCK during the research and development stage. To address this, we introduce an advanced ensemble machine learning technique designed to estimate the binding affinity between molecules and LCK. This comprehensive method includes the generation and selection of molecular fingerprints, the design of the machine learning model, hyperparameter tuning, and a model ensemble. Through rigorous optimization, the predictive capabilities of our model have been significantly enhanced, raising test R2 values from 0.644 to 0.730 and reducing test RMSE values from 0.841 to 0.732. Utilizing these advancements, our refined ensemble model was employed to screen an MCE -like drug library. Through screening, we selected the top ten scoring compounds, and tested them using the ADP-Glo bioactivity assay. Subsequently, we employed molecular docking techniques to further validate the binding mode analysis of these compounds with LCK. The exceptional predictive accuracy of our model in identifying LCK inhibitors not only emphasizes its effectiveness in projecting LCK-related safety panel predictions but also in discovering new LCK inhibitors. For added user convenience, we have also established a webserver, and a GitHub repository to share the project.

TransFoxMol: predicting molecular property with focused attention.

Predicting the biological properties of molecules is crucial in computer-aided drug development, yet it's often impeded by data scarcity and imbalance in many practical applications. Existing approaches are based on self-supervised learning or 3D data and using an increasing number of parameters to improve performance. These approaches may not take full advantage of established chemical knowledge and could inadvertently introduce noise into the respective model. In this study, we introduce a more elegant transformer-based framework with focused attention for molecular representation (TransFoxMol) to improve the understanding of artificial intelligence (AI) of molecular structure property relationships. TransFoxMol incorporates a multi-scale 2D molecular environment into a graph neural network + Transformer module and uses prior chemical maps to obtain a more focused attention landscape compared to that obtained using existing approaches. Experimental results show that TransFoxMol achieves state-of-the-art performance on MoleculeNet benchmarks and surpasses the performance of baselines that use self-supervised learning or geometry-enhanced strategies on small-scale datasets. Subsequent analyses indicate that TransFoxMol's predictions are highly interpretable and the clever use of chemical knowledge enables AI to perceive molecules in a simple but rational way, enhancing performance.