Artificial Intelligence In Chemistry: Practical uses of AI
Will our daily task of trying to find new synthesis routes, molecules and catalysts be over once AI is fully integrated within chemistry research? I think your job is safe.. For now. You’d have to have been living in a lead walled lab for the century to not stumble on an article about AI or machine learning in chemistry. Here we’ll list the most practical applications of AI in Chemistry.
“It’s not an AI article without a terrible graphic” – AutomationChemistry.com 2021
Reaction Prediction Analysis
Apart from those final year undergraduate students who’ve just sweated out revising for their final organic exams, it’s not practical or worthwhile to retain a large library of synthesis routes in your head (sorry undergrads) or painstakingly raid through the library of hefty textbooks to make viable predictions of reaction outcomes. This is one of the applications of AI in chemistry to predict these for us. For example RXN by IBM (guide here) you can draw in your reagents and converting it in to a SMILES language it will trawl through publications to make up similar combinations to make a prediction along with ranking its confidence level. So essentially it converts our scribbles into a legible language and then tries to make sense of it, thanks technology. Give RXN ago and see for yourself if you think the predictions are any good.
Retrosynthetic Analysis
Similar to the predictive analysis of a forward reaction, AI is being used to create frameworks for detecting retrosynthetic pathways to molecules from publications and combining machine learning. This again is going to convert the inputted molecule into the written format see example below. it will then be able to predict the areas likely for cleavage to occur and back this up with routes documented in it’s source literature.
Most programs will give a variety of different routes representing the usual process in retrosynthetic analysis and even filter to avoid some starting materials to make the analysis fit your requirements more.
Drug Discovery
Medicinal chemistry provides much force for advancements in modern day chemistry not just in the whole ‘curing the world’ mission but also the money it brings. AI is already heavily involved in drug discovery and a group from MIT screened more than a hundred-million chemical compounds in just a few days for their abilities to kill bacteria effectively. “Using a machine-learning algorithm, MIT researchers have identified a powerful new antibiotic compound. In laboratory tests, the drug killed many of the world’s most problematic disease-causing bacteria, including some strains that are resistant to all known antibiotics. It also cleared infections in two different mouse models.” Read more on this story here.
Due to the vast scale of screenings required for drug discovery it’s no wonder why AI will help accelerate research and help find the most efficient and effective drug candidates. As you can imagine the AI systems that are carrying out these complex and vast predictions are not free to use or open source.
Prediction and Detection of Molecular Properties
From coatings, polymers and many other material science applications, the properties are always crucial for the pursuit of new molecules for their unique applications. With the help of AI predictions can be made without having to tirelessly synthesise each closely related molecule and test it’s properties by hand (or robot), these can be done AI predictions. Not only does this save time but a much larger scope can be tested for a wide range of applications to discover new molecules and have a good understanding of the properties of these before even attempting to synthesise.
AI and machine learning are only going to become more and more common place in research and industry across the chemical world and the opportunities and developments it will bring with it will for sure be beyond what was before perceived as possible.
Read a recent nature review on machine learning and AI here
Leave a Reply