Comparison of QSAR and Molecular Similarity Approaches for a Structure-Activity Relationship Study of DHFR Inhibitors. DHFR Inhibitors: QSAR and Molecular Similarity Approaches

A comparison is made between classical QSAR approaches using multivariate linear regression analysis and molecular similarity analysis (MSA) using inhibitors of dihydrofolate reductase (DHFR). A wide variety of descriptors derived directly from molecular structure are examined with both QSAR and MSA approaches. Three separate measures of molecular similarity are examined to determine how each compares with the QSAR method in the context of the descriptors used. The QSAR method yielded a model with a R2 = 0.927 whose descriptors indicate that molecular size, shape and hydrogen bonding acceptor groups are important to activity. Molecular similarity approaches did not provide as good a model.The best correlation between pK(i) values and any given similarity metric was obtained with the Tanimoto coefficient.