A new reinforcement learning system has been created at the University of North Carolina School of Pharmacy. The system is called the Reinforcement Learning for Structural Evolution, also known as ReLeaSE for short. It can teach itself to design drug molecules from scratch which has never been seen before in the field of AI. There is vast potential with a program like this that can not only create new drugs to fight everything from the common cold to cancer but can also help accelerate the process by which these drugs get through human trials, approved by the FDA, and onto the market.


ReLeaSE works by utilizing a novel algorithm that is comprised of two neural networks which work together and have a mutual relationship. One of them has been programmed with syntax and linguistic rules that exist as the basis for chemical structures. The other learns how to combine this data into a working chemical model which is then used as a blueprint for the design of a new drug. The first network constantly monitors the other one to make sure the learned combination follows the rules that it was programmed to follow to ensure accuracy.

The program is a very powerful innovation to virtual screening which is the method that is currently in widespread use in the pharmaceutical industry today to identify viable drugs. This team has used ReLeaSE to generate specific sets of molecules with a certain degree of bioactivity and safety profiles. We have no doubt that the resulting drugs will be safe in the sense that they won’t directly be toxic to the human body but as we will discuss in the next section safety should go far beyond that.


This new system is revolutionary but nothing is without its potential drawbacks. The first thing that comes to mind is the safety profile of the drugs it creates. After all, that is one of the major issues plaguing the pharmaceutical industry at the moment. Prescription drugs typically have a questionable safety profile.

They are able to treat one ailment by introducing one or two major chemicals into your system only to potentially cause a number of unwanted side effects. The creators of this project claim that rules that are designed to ensure safety have been programmed into the first of the two neural networks. However, when one considers just how difficult it can be to accurately predict the way in which a drug will affect someone and because our bodies are incredibly complicated then this programs ability to make these drugs safe should be questioned. But at the same time it isn’t out of the realm of possibility that the programmers who created this system will make algorithmic improvements that solve this problem. We must hope that this will be the case.


The above graphic highlights to what extent the Healthcare industry is being transformed with Artificial Intelligence technology.


There is another issue with AI programs of this kind. Currently we have a major problem in the pharmaceutical industry that is characterized by creating drugs that aren’t really designed to cure you at all. Instead these drugs are designed to keep you dependent upon long term in order to generate revenue.

Just a couple of examples of this are medications like Suboxone and Methadone. Prescribing these two drugs to a heroin addict in lieu of treatment is equivalent to trading an illegal addiction for a legal one. Ethical issues like these are beyond the scope of this post but it is important to briefly touch on this because when you introduce an AI into the equation that can create very powerful drugs from scratch at a very fast rate there is potential for abuse. The optimistic point of view of course is that this program will be used for the benefit of humanity in its pursuit of cures for cancer and Alzheimer’s and Parkinson’s and other major diseases. The alternate possibility is that programs of this nature may be used to create stronger and more addictive drugs such as narcotics that only serve malicious purposes.


Given our society’s dark history when it comes to designer medications one might be weary or overly pessimistic when thinking about the future of this kind of technology. The fact of the matter is AI is here to stay, and every single industry including the pharmaceutical industry will experience a complete overhaul regarding not only their production rate but their production methods as well. There will always be potential for misuse just like with any new technology. The important thing is that we are aware of these immoral intentions and seek to guard ourselves from their implementation and to also be aware of any problems and seek to find a solution. Overall the impact of this technology will be positive when its all said and done.

Leave a Reply

Your email address will not be published. Required fields are marked *