Construction and validation of QSAR models for thiazole derivatives with antifungal activity

Cryptococcosis is an infectious disease with worldwide distribution and clinical relevance, mainly caused by C. neoformans and C. gattii. Cryptococcal meningitis is one of the most important HIV-related opportunistic infections and it is associated with significant mortality (~600,000 deaths/year).1 Aiming to develop new substances with antifungal activity against C. neoformans, 2D and 3D-QSAR studies were carried using a series of 29 thiazole derivatives with determined minimum inhibitory concentration (MIC). The 2D structures were constructed and a conformational analysis generated the lowest energy conformations and, after, they were aligned according common scaffold. Antifungal space (pMIC for C. neoformans inhibition), drug-like space (molecular weight, ClogP, number of hydrogen bond donors and acceptors, fraction of sp3 carbons, and number of rotatable bonds), and structural space (MACC fingerprints) were considered for individual hierarchical cluster analysis (HCA) which guided training and test sets (4:1) separation. Then, 2D (HQSAR) and 3D (CoMFA and CoMSIA) QSAR models were constructed (Sybyl X 2.1) by using different parameter combinations and internally validated. After generation and comparison of their robustness, the five most robust models were selected and submitted to external validations, according to OECD guidelines. As a result, all selected models presented higher Q2(F1), Q2(F2), Q2(F3) and CCC values (>0.6) than other models, showing their acceptable predictive power. Therefore, the three employed techniques as well as constructed models could be employed to predict the biological activities of new proposed substances in the design of new antifungal agents.