Drug synthesizing machines are one thing, but what if a machine could invent a new drug from scratch?
That’s what they’re doing at the University Of North Carolina at Chapel Hill. The concept is called Reinforcement Learning for Structural Evolution (ReLeaSE). The system designs drug molecules and has the potential to rapidly accelerate the development of new drugs.
While humans are more intelligent than computers, supercomputers can work around the clock at incredible speeds evaluating various possibilities (and maybe even try them out, in the case of some rapid prototyping machines i’ve seen), and humans can then assess the generated drug formulas for their safety and do all the other necessary work required to ensure that they are safe.
Unsurprisingly, the ReLeaSE system is powered by a computer program with an algorithm that utilizes two neural networks. A neural network is a group of neurons that communicate with each other. The human brain utilizes neurons, and artificial neural networks (ANN), which are sometimes utilized by projects such as these attempt to simulate that neuronal behavior.
One of the two neural networks in ReLeaSE acts somewhat like a teacher, and the other acts as a student. The teacher is equipped with knowledge of 1.7 billion biologically active molecules, and the student not only learns, but proposes molecules that may be useful for new medicines.
Considering the technology available today, this may one day be a part of a larger system that sends its output to a rapid prototyping machine to automate the drug generation process further.