Molecular connectivity and substructure analysis.

Antimicrobial and antiviral data sets were analyzed by molecular connectivity. Standard structure--activity relationship equations of high quality were produced in both cases. For phenyl propyl ether activity against Staphylococcus aureus, the three variables 1chi, 3chiP, and 4chiUPC yielded an r of 0.957, significantly better than a pi,sigma analysis. Analysis of benzimidazole antiviral data (Lee strain, B flu virus) revealed that the one variable, 6chiP, yielded an r of 0.950, also better than a reported Hansch analysis. Both data sets were further analyzed by partitioning the important regression variables into terms representing various structural features of the molecules. For the phenyl propyl ethers, the para-region of the phenyl ring is important for improved activity and the negative coefficient on 3chiP corresponds to decreased activity for vic-dihydroxy compounds. For the alkylbenzimidazoles, substitution on the five-membered ring is highly important. No discrimination of six-membered ring positions was revealed. These structure--activity relationship observations can form the basis for synthetic decisions to improve activity.