Abstract

PREDICTION OF NOVEL ANTI-EBOLA VIRUS COMPOUNDS UTILIZING ARTIFICIAL NEURAL NETWORK (ANN)

Ronald Bartzatt*

Abstract

Artificial Neural Network (ANN) analysis is shown to predict the molecular properties of new anti-EBOLA compounds following training/learning by use of 60 previously known and studied drugs. Following training/learning by applying properties of 60 known drugs the TIBERIUS ANN system can efficiently predict the molecular properties of comparable new drugs. Molecular weight (MW) is an important and dominant property of perspective drugs considered for clinical trials. TIBERIUS ANN was able to predict comparable values of MW for drugs following training cycles. One-way ANOVA, F and T tests indicate that actual and predicted MW have the same means (P=.99). Passing-Bablok regression showed that ANN predicted MW are comparable to actual MW. The coefficient of variation indicated actually less variation in predicted MW as opposed to actual MW. A plot of actual MW values versus ANN predicted MW values, produced a line having no departure from linearity (P=.82), and a 95% ellipses having 55 drugs therein. TIBERIUS ANN allows investigators to input separate property values to predict suitable outcome based on the 60 known drugs. ANN prediction of pharmaceutical properties of new drugs is shown to be efficient and accurate when based on a known set of drugs for training/learning cycle.

Keywords: Ebola, Artificial Neural Network, Viral Hemorrhagic Fever.