PLS modelling of structure—activity relationships of catechol O-methyltransferase inhibitors

Summary

Quantitative structure-activity analysis was carried out for in vitro inhibition of rat brain soluble catechol O-methyltransferase by a series (N=99) of 1,5-substituted-3,4-dihydroxybenzenes using computational chemistry and multivariate PLS modelling of data sets. The molecular structural descriptors (N=19) associated with the electronics of the catecholic ring and sizes of substituents were derived theoretically. For the whole set of molecules two separate PLS models have to be used. A PLS model with two significant (crossvalidated) model dimensions describing 82.2% of the variance in inhibition activity data was capable of predicting all molecules except those having the largest R₁ substituent or having a large R₅ substituent compared to the NO₂ group. The other PLS model with three significant (crossvalidated) model dimensions described 83.3% of the variance in inhibition activity data. This model could not handle compounds having a small R₅ substituent, compared to the NO₂ group, or the largest R₁ substituent. The predictive capability of these PLS models was good. The models reveal that inhibition activity is nonlinearly related to the size of the R₅ substituent. The analysis of the PLS models also shows that the binding affinity is greatly dependent on the electronic nature of both R₁ and R₅ substituents. The electron-withdrawing nature of the substituents enhances inhibition activity. In addition, the size of the R₁ substituent and its lipophilicity are important in the binding of inhibitors. The size of the R₁ substituent has an upper limit. On the other hand, ionized R₁ substituents decrease inhibition activity.