They say clinical research of a drug costs as much as a passenger airplane, and therefore there are strong requirements to use new technologies that can not only substantially reduce costs but also make the development processes time efficient. Now, researchers from the University of North Carolina have formulated a novel approach using artificial intelligence (AI) that can self-teach the designing of new drug molecules and may be able to radically accelerate the development of new drug candidates.
Named as Reinforcement Learning for Structural Evolution, or in short ReLeaSE, the AI system leverages computer programming and algorithms that includes two neural networks functioning as a student and a teacher. The teacher will be held responsible for linguistic and syntax rules pertaining to the formulations of chemical structures from 1.7 million predefined active molecules whereas the student acquires information overtime and evolves as a promising molecule that can be utilized as a new medicine.
ReLeaSE promises to pave way to an innovative method of virtual screening, akin to the computational methods that are currently being widely used by the pharmaceutical companies aspiring to identify novel drug candidates. With ReLeaSe, researchers will be able to analyze among the volumes of chemical libraries and identify between specific bioactivity and safety requirements. The system is capable of designing molecules appropriate for physical properties such as solubility into water, melting point, and formulating new biochemical compounds.
The method is expected to be a boon for the pharmaceutical industry that is consistently aspiring for a new approach that can reduce the time that is currently required to pass new drugs through clinical trials.