Multivariate Linear Regression Models Based on ADME Descriptors and Predictions of ADMET Profile for Structurally Diverse Thermolysin Inhibitors

During development a large number of molecules are rejected as drug candidates due to unfavorable ADMET (absorption, distribution, metabolism, excretion and toxicity). In Silico ADMET predictions early in development may shorten drug discovery timelines. QSAR modeling of relationships between molecular descriptors (MDs) and activity profiles is a well known approach for predicting ADMET properties. Inhibitors of thermolysin and other thermolysin like zinc-metalloproteases are promising as therapeutic compounds. In the present paper ADMET parameters of 25 thermolysin inhibitors were predicted using QSAR models derived from multiple linear regressions (MLR) analysis and artificial neural network using the web-based tool PreADME version 1.0 (http://preadmet.bmdrc.org/preadmet/index.php). The QSAR models indicated that best correlation was obtained between polar surface area (PSA) and the MDs. PSA is important for bioavailability of the compound. The model had a regression coefficient (R2) of 0.915 (p<0.001).