In Silico Classification Models for Anticancer Drugs

In spite of extensive research in the understanding of neoplastic diseases, the success rate for oncology drugs continues to be very low. Accordingly, a major challenge before the scientific community is to design new chemical entities that will be highly selective for cancer cells so as to minimize side effects. An urgent need for prompt adaptation and systematic utilization of In silico approaches in drug design/research has received wide acceptance due to its potential in accelerating drug discovery process with improved efficiency. Amongst In silico approaches, classification models play a prominent role in prediction of the biological properties of newly designed compounds before their synthesis and to prevent non-optimal use of resources. These models can be derived from either in vitro or in vivo assay data and can be subsequently utilized for better understanding/extrapolation of data in terms of desired biological activity. Successful validation of the said predictive model(s) leads to swift cycle times, minimization of animal sacrifice and an early indication of drug attrition/failure amalgamated with reduced cost. Though correlation models far outnumber classification models for development of various therapeutic agents but the significance of classification models for development of anti-cancer agents cannot be underestimated. Various machine learning techniques employed for development of classification models for anti- cancer activity have been briefly reviewed in this chapter.

Keywords: Anticancer agents, classification modeling, classification techniques, decision tree, molecular descriptors, moving average analysis, random forest.