A Computer System for Predicting Antimycotic Properties

Abstract

The paper discusses a computer system for predicting antimycotic properties, which can predict a certain compound’s toxicity and useful properties based on its chemical structure, as well as find similar synthesis parameters from a compound synthesis methods database. The system contains two neural networks (to predict toxicity and useful behavior from molecular descriptors and the SMILES notation), a rule base for synthesizing novel antibiotics, as well as an antimycotic compound synthesis database. The system can help significantly lower the development cost for novel antibiotics. It has been tested using antifungal polyene macrolide antibiotics.