Chemistry goes digital: The Chemputer

20 July 2018
Posted by Marika Ottman.

A robot powered by Artificial Intelligence (AI) is helping chemists to discover new molecules and chemical reactions...

Mixing chemicals together causes reactions and produces new molecules. But with so many different chemicals in existence, the possible combinations are endless, and there are millions of chemical reactions yet to be tried.

Doing these tests, though, is very time consuming, laborious and has a low success rate; it’s also very susceptible to human failings, like bias and fatigue. To surmount this, University of Glasgow chemist Lee Cronin and his team have invented a robot to test chemical combinations for them.

“I needed to make a system that would do chemistry without bias, and go and discover with no expectation of discovery. So I made a robot that did all of the boring stuff,” jokes Cronin.

Cronin’s solution is a robot chemist. Although one might imagine a robot walking around in a lab coat, this robot works quite differently.

The system consists of a series of computer-controlled pumps and valves in a fume hood, connected to various devices including spectrometers that can test the molecular properties of the results. Chemicals are administered into the system and the computer determines which chemicals are combined for testing. Subsequent reaction choices are informed by learning from previous experiments it's conducted.

The device uses its analytical equipment to probe the molecular bonds, the masses of the resulting compounds, and the magnetic resonance properties of the molecules. This combination of characteristics is used to determine if a new molecule, possibly produced in a new way, has been discovered.

In a pilot test, Cronin and his team provided the robot with a menu of 18 reactive chemicals and tasked it with testing 1000 different combinations. The robot tested the first 100 combinations at random, and used the results of those chemical combinations to predict the outcomes of the remaining 900 combinations with 80% accuracy.

The robot’s ability to learn and make scientifically-informed decisions makes it useful in discovering new molecules as efficiently as possible. And because the robot operates independently of any existing databases, it can explore "chemical space" outside of the confines of standard organic synthesis methods. So far, the robot chemist discovered 4 new molecules, and rediscovered numerous already-documented molecules. 

"This approach is a key step in the digitisation of chemistry," explains Cronin. “It will allow the real time searching of chemical space leading to new discoveries of drugs, interesting molecules with valuable applications, and cutting cost, time, and crucially improving safety, reducing waste, and helping chemistry enter a new digital era..."


Add a comment