Nucleo-amino acid derived AMO as a potential JAK inhibitor: machine learning screening, docking, molecular dynamics, ADMET/toxicity, and MM-PBSA analysis

L-amino acids and nucleotide bases are essential biomolecules in all living organisms and represent attractive building blocks for designing safer, more effective drugs. In this study, we designed JAK kinase inhibitor candidates derived from these natural compounds and evaluated them using an integrated computational workflow. MarvinSketch was used to generate 100 structures, which were optimized in Spartan; frequency calculations confirmed the absence of imaginary modes and enabled calculation of key physicochemical properties. The candidates were then screened against five known JAK inhibitors using machine learning models (SVR, RF, KNN, and an ensemble approach), yielding (3S,5S)-5-((2-amino-6-oxo-1,6-dihydro-9H-purin-9-yl)methyl)-3-(4-aminobutyl)morpholin-2-one (AMO) as the top hit. Molecular docking identified favorable active-site interactions, and Abrocitinib (ABR) was selected as the closest comparator with the best overlay to AMO. Molecular dynamics simulations showed stable AMO binding, supported by RMSD/RMSF, radius of gyration, hydrogen-bonding, and DSSP analyses. PCA indicated a more compact conformational distribution for AMO than ABR, consistent with reduced large-scale motions, and DCCM supported favorable correlated dynamics. Pharmacokinetic profiling and ProTox-3.0 toxicity predictions classified both compounds as class 4, but AMO exhibited substantially lower predicted toxicity than ABR. Finally, MM-PBSA calculations yielded negative binding energies, confirming that the AMO–protein interaction is thermodynamically favorable and strong.