Cytoprotection and cytotoxicity data of anti-HIV 2,3-diaryl-1,3-thiazolidin-4-ones were subjected to QSAR study using Fujita-Ban type analysis and a mixed approach based on Hansch and Fujita-Ban analyses. Apart from appropriate indicator and integer variables encoding different group contributions, different physicochemical variables like hydrophobicity (p) and steric (molar refractivity) parameters of aryl ring substituents of the compounds were used as predictor variables. Furthermore, Wang-Ford charges of the common atoms of the compounds calculated from molecular electrostatic potential surface of AM1 optimized geometries of the compounds and various topological parameters were used as additional descriptors. The variables for the multiple regression analyses were selected based on principal component factor analysis, and generated equations were statistically validated using leave-one-out technique and predicting the activity data of test set compounds. The statistical qualities of the equations for cytoprotection data (explained variance ranging 64.5 -80.3%, leave-one-out predicted variance ranging 44.3 -59.4%) are better than those for cytotoxicity data (explained variance ranging 59.7 -60.6%, leave-one-out predicted variance ranging 52.4 -54.4%). The analysis explores the structural and physicochemical requirements of the compounds for cytoprotection and cytotoxicity.