Development and Evaluation of Novel Molecular Descriptors for Accelerating Drug Discovery Process

Abstract

Historically man looked at herbs for alleviating pain and suffering. It is not surprising that throughout history man has searched for remedies to fight against disease. Ancient civilizations had complete treatises as compilation of herbs or mixtures used as therapeutic substances to alleviate and treat disease (Enrique, 2011). Modern day remedies do not depend entirely on herbs but on compounds that evolve through a complex, multivariate drug discovery process for economical production of a stable, potent compound with high therapeutic efficacy and low toxicity. In general drug discovery (DD) process consists of seven sequential steps: selection of target disease, target hypothesis, lead molecule identification, lead optimization, pre-clinical study and clinical trial and pharmacogenomic optimization. Generation/ optimization and testing lead chemical compounds have been considered the main bottlenecks in drug discovery process. newlineIn the classical approach, a number of chemicals in the range 103 104 is synthesized and assayed to identify lead compounds, whereas the introduction of combinatorial chemistry (CC) and high throughput screening (HTS) has increased the range to 105 106. Although automation has increased the speed and reliability of HTS, still these chemical compounds need to be synthesized before the screening. Machine learning models (ML) are useful to enhance lead identification using a virtual HTS (vHTS) screening technique even in the range of 107 108 molecules as there is no need to synthesize all these compounds. The ML models are also used to predict molecules that have a high probability of interacting with the selected target. In the vHTS approach chemical compound structures are generated using computer programmes that offer possibility to explore a diverse chemical skeletons and a much higher number of molecules (Ivanciuc, 2008). Increased efficiency of drug discovery process owe to computational screening techniques which are faster, simple and less expensive than HTS in identification of ...