A homology model of human 3β-HSD type 2 has been developed from homology modeling techniques using Phyre2 server and refi ned by ModRefi ner. The PROCHECK, QMEAN and ProSA-web online tools were carried out to evaluate the stereochemical quality of the model. The Ramachandran plot resulted from PROCHECK showed that 84.5% residues are in the most favored region, 13.7% are in the additional allowed region, 1.5% are in the generously allowed region and 0.3% are in the disallowed region. The QMEAN (Z-score) are 0.509 (-3.006) and Z-score of ProSA-web is -7.10. The negative values of protein fold energies also found in almost all sequences. Furthermore, molecular docking was also applied to validate the model using MOE. The hydrogen bonding interactions with Tyr154, Ser124, and Ser218 are found in all docked substrates as well as known inhibitors (trilostane and epostane). A dataset of azasteroid inhibitors were also docked into the substrate active site of human 3β-HSD2. These docked structures were utilized to construct corresponding docking-based QSAR equation by employing genetic algorithm (GA) statistical analysis. The contructed best QSAR equation has a robust predictive power according to its statistical parameters, hence may be applied to supersede the default scoring function provided by docking software. These results indicate that the human 3β-HSD2 model was successfully evaluated as a good model.