Machine learning and docking-based insights into descriptor, activity, and chirality cliffs of SYK inhibitors.

Spleen tyrosine kinase (SYK), a critical non-receptor tyrosine kinase, is a significant target in therapy for cancer, autoimmune diseases, and inflammation. This study combines machine learning with molecular docking to overcome challenges related to activity and chirality cliffs in SYK inhibitors. A robust and statistically reliable multiple linear regression model (R2tr = 0.799, Q2LMO = 0.796) was developed, complemented by docking analyses of 1854 molecules. The findings reveal that inhibitory activity is governed by specific structural features, including the solvent accessible surface area of certain H-bond donors, the presence of sp2-C, and various nitrogen atom types. Crucially, the configuration of chiral centers significantly influences the activity profile. This dual approach effectively decodes the activity and chirality cliffs, offering valuable insights for the future optimization and development of potent SYK inhibitors.