Integrated machine learning, pharmacophore screening, and structure-based computational approaches for the identification of novel JAK3 inhibitors.

AIM: Janus kinase 3 (JAK3) is a key therapeutic target in autoimmune and inflammatory diseases. This study aimed to identify novel JAK3 inhibitors using an integrated machine learning and structure-based computational approach.

MATERIALS AND

METHODS

Random Forest Regression model was developed from 4,336 JAK3 inhibitors using molecular descriptors and Morgan fingerprints, then applied to screen 4.6 million Enamine library compounds. Top hits underwent pharmacophore screening, molecular docking, triplicate molecular dynamics simulations, and MM-PBSA binding free energy calculations.

RESULTS

The optimized model demonstrated strong predictive performance and reliable pIC50 estimation. Docking analysis further prioritized the lead compounds, and MD simulation triplicates confirmed stable binding of prioritized compounds within the JAK3 active site, with persistent interactions involving hinge residues Glu903 and Leu905, and charge residues Cys909, Arg911, and Arg953. Lead compounds (Z2953646389, Z2235801948, and Z56791230) were stabilized by Coulombic and van der Waals interactions, exhibiting more favorable MM-PBSA ΔG values than the reference inhibitor ritlecitinib.

CONCLUSION

Integrating machine learning with structure-based computational approaches provides an efficient strategy for JAK3 inhibitor discovery. The identified leads show strong predicted potency and favorable binding profiles, warranting further experimental validation.