Multiple Logistic Regression Modeling of Compound Class as Active or Inactive Against COX-2 and Prediction on Designed Coxib Derivatives and Similar Compounds

Abstract

The development of next generation non-steroidal anti-inflammatory drugs (NSAIDs) is one active area of research as inflammatory diseases continue to afflict over 1.5 billion people worldwide. The publicly available and computationally accessible chemical and biological data provide a wellspring of information for any research pursuit that could expedite the discovery of new anti-inflammatory drugs. Computational statistics is a handy tool in establishing quantitative relationship between the anti-inflammatory activity and the key molecular features that determine the compound’s medicinal property. In this work, Multiple Logistic Regression (MLogR) was employed to develop a mathematical model of the inhibitory activity of a compound on cyclooxygenase-2 (COX-2), an enzyme that facilitates the production of inflammatory prostanoids. The best model with hit ratio of 94% and 91% on the train and test set, respectively, was used to predict the classification (i.e. active or inactive) of newly designed coxib Derivatives and Similars obtained through similarity search. The predicted actives were further screened based on their quantitative estimate of druglikeness (QED), synthetic accessibility, and ADMETox properties. The selected top 15 hits have superior confidence as actives, are highly druglike and easy to synthesize, and generally possess outstanding drug profile.