AI in medicinal chemistry

We are always surrounded by medicine, whether this be through, for example, the cabinet in your house containing prescription drugs or walking by a pharmacy during the day. It is no secret that medical drugs are essential - they both mitigate the symptoms of disease and even prevent further future illness. However, whilst ingesting a tablet is easy for most, it seems to be that we can sometimes forget the vigorous amount of scientific research that goes into successfully synthesising a new drug, i.e., the core of medicinal chemistry. This process typically takes up to an astounding 10 years or more, but with new artificial intelligence (AI) emerging it is thought to be that this number will lower.

What exactly is artificial intelligence? It can broadly be defined as the ability to produce human intelligence through the use of machinery such as computers or software. Based on this, one may question why AI is needed if we can just simply communicate ideas through writing, speaking and so on. The answer is increased efficiency – one example of man made neurones is discussed on the website Interesting Engineering, which are able to produce impulses up to one billion times per second. Fascinatingly, this is quicker than humans, so it could also be argued that AI is actually better than us!

There are many phases of the drug development process, from early pre-clinical research to post-market surveillance. When a drug is administered, the body uses enzymes such as mainly those from the CYP family to break the compound down into smaller structures, through a process known as metabolism. Drug metabolism can create toxic molecules that are able to covalently bind to proteins in the body causing serious illness, but also molecules that can be harmlessly excreted through faeces or urine. Of course, chemists can look for sites of metabolism by studying the angles and positions of atoms, however AI is able to do this much quicker and with higher accuracy.

SuperCYPsPred is an example of a free online web application that can predict if a drug may be a CYP enzyme inhibitor in pre-clinical drug discovery, as the software is able to identify five of such inhibitors. Through this, we can understand how a drug’s metabolic pathway may differ and investigate further early on, allowing scientists to make structural changes before proceeding onto the next phase of development. Through this, millions of pounds can be saved from marketing an unsuccessful drug as well as decrease the chances of causing injury to the public.

AI is also able to use machine learning (ML) to carry out tasks. ML is when machinery processes a large data set and identifies complex patterns to problem solve. From this then comes deep learning (DL), which allows this ML to be applied in different fields. For example, DeepCE is a “novel deep learning computer model” that helps predict changes in gene expression with certain drugs. It is able to do this by using the following two sources: DrugBank which contains data for 11,000 safely approved drugs and the L1000 dataset that has information on over 1 million perturbed organ tissue gene expressions. From this, researchers were able to obtain 10 drug candidates for the treatment of COVID-19 infection, in which 2 have been successfully marketed.

Based on the above, it is clear that AI holds a lot of power in speeding up the drug discovery and development process. With the technology sector advancing in general as well, we are looking at a future where AI will become even more dominant in the pharmaceutical research industry. Whilst AI can predict several drug properties, it is also important to remember that we physically cannot predict every single thing out there – we can only try our best, which AI is aiding.

Written by Harsimran Kaur

Related articles: AI in drug discovery / A breakthrough procedure in efficient drug discovery / Role of chemistry in medicine