Mark Warne is the director of a company that has the ambition of digitising chemistry. Deepmatter aims to do this by combining chemistry with technology, both hardware and cloud-based software, to provide chemists with the tools to make discoveries more productively and cost-effectively. Essentially, Deepmatter is automating the data capture of chemical reactions and creating a library of easy-to-replicate reactions.
The image of a chemist is a person in a lab coat surrounded by petri dishes, dropping solutions from pipettes into test tubes to create fizzy and smoky reactions. This is not a million miles from the truth with chemists creating reactions by mixing together reagents to produce something that has some utility in an industry, such as a drug molecule. The reagents, or ingredients, are mixed, stirred, heated and agitated in a scientific environment.
The chemists are usually following a set of instructions to get a specific result, usually from another chemist who has done that same reaction in the past. But according to Warne, there is a problem here. He said: “The ability for me to articulate a piece of science to you with some degree of certainty is difficult because of the level of detail that is required. This means it is hard for you to be certain that you are reproducing accurately.”
Warne said that this has led to a “reproducibility crisis” with there being just a one in four chance of reproducing an experiment that has been done by somebody else.
But, the hardware devices created by Deepmatter can collect data – including images, sound, temperature, as well as ultra violet and infrared readings - from the beginning to the end of a reaction. According to Warne: “This level of data fidelity allows you to understand not just whether a reaction was a success or not, but where it was successful, why it was successful, and what set of actions made a reaction capricious or not.”
After the reaction, this data laid out as a set of instructions and put into libraries, almost like recipes in a cookbook, and lets others know how to recreate a similar reaction in the future.
In addition, during the actual reaction, the sensors that are picking up the data can also detect a deviation in the process from the intended course of action. A feedback loop is created to ensure that what is happening during the reaction is what the chemist intended. If something is not going to plan, the sensors take note and take action.
Warne gave the example of stirrers, which are often magnetic, that could flip or jump, and stir the solutions of a reaction inefficiently. “Unless you are watching the reaction, you wouldn’t know that. However, by using the sensors, you are able to determine the set speed of a stirrer and the speed that the device is traversing at. If there is a lack of correlation between the speeds then you can automatically stop the stirrer and completely start it again,” said Warne.
With these libraries and feedback loops, Deepmatter is using data science to aid the science of chemistry.