Qsar study for antibacterial activity from molecular structures of benzamide and oxazolidinone derivatives / Noor Hidayah Naemat

Quantitative structure-activity relationship (QSAR) approach is one of fields in computational chemistry. There are many successful predictive models developed for various activity predictions. Due to the emergence of new patterns of resistance of bacteria to antibacterial agents, new antibacterial agents are needed. Data of benzamide and oxazolidinone derivatives from previous studies were reanalyzed for their antibacterial activities by using different descriptors generated by DRAGON 5. Two methods of variables selection were used, which are stepwise regression and forward selection procedures available in MINITAB 14 statistical software. Multiple linear regressions (MLR) analysis was used in developing QSAR models to determine whether the descriptors used can give good QSAR model. The QSAR models have been evaluated and validated to determine their stabilities and prediction capabilities. Six QSAR models have been developed and their statistical results were compared with data from the previous studies. Four QSAR models developed have higher correlation coefficient, R2 and cross-validation R2 V values, showing higher stabilities and prediction capabilities. The best QSAR model has R2= 0.93 and R2 V= 0.91 and three descriptors were included in this QSAR model. The R2 V for -log MIC of B. subtilis is 0.912, S. aureus is 0.710, E. coli is 0.766, MIC of A. fwcum is 0.272, A. paraciticus is 0.789, and log (1/C) of S. aureus is 0.692.