Abstract

PREDICTION OF MOLECULAR PROPERTIES OF NOVEL COX-2 INHIBITORS UTILIZING ARTIFICIAL NEURAL NETWORK ANALYSIS

*Ronald Bartzatt

Abstract

Selective COX-2 inhibitors are a category of nonsteroidal antiinflammatory drug (or NSAID) that directly and selectively targets cyclooxygenase-2 enzymes, which are responsible for pain and inflammation. The development of various NSAID type drugs to selectively inhibit COX-2 enzymes have been found to be very effective in suppressing certain serious pathological conditions such as neurodegenerative conditions, certain cancers, as well as pain. Neural networks are a nonlinear method of modeling extremely complex functions. This study shows the effectiveness of artificial neural network analysis (ANN) of the molecular properties of already proven COX-2 targeting NSAID’s, to predict and model new drug structures. In this study, the molecular properties of seven known COX-2 inhibitors are utilized to produce a neural network model for determining molecular properties for new COX-2 drugs. Important properties such as Log P, polar surface area, molecular volume, etc. are input to the model to establish learning parameters for predicting new properties. The predicted molecular weight of new drugs are statistically alike to the molecular weights of the training parameters (F and t test: P=0.069). Properties of new COX-2 inhibitors with the range in value, include the following: Log P (from 0.706 to 4.4477); molecular weight (from 275.26 to 411.65); and molecular volume (from 260.0 Angstroms3 to 317.01 Angstroms3). The graphical comparison of ANN predicted properties for new COX-2 drugs are presented. ANN learning model is shown to be useful in the prediction of molecular properties for potential new COX-2 inhibitors.

Keywords: ANN, artificial neural network, COX-2, TIBERIUS.