Self-driving laboratories: a reality waiting to happen
Imagine a situation where, given a request, a lab automatically chooses what experiments to do, robotically carries this out, tracks the reaction with integrated sensors, acquires and analyses the results, and then decides what experiment to do next. This seems like science-fiction, but certain truths may be closer than you think.
Many terms have been coined to represent this idealised future of experimentation: 'Self-driving-labs', 'Chemputer', and 'The Artificial Chemist' to name a few. This concept of removing all human elements in physical experimentation may be horrifying to some or very attractive to others; this full reality is a long way off, but certain technology innovations are here today and can have an immediate impact.
IDTechEx has identified three core technology pillars that are required. Laboratory informatics, materials informatics (or cheminformatics/bioinformatics as appropriate), and robotics. IDTechEx has released a detailed report about materials informatics, detailing the key technologies, players, applications, and market outlook. For more information see 'Materials Informatics 2020-2030'.
Figure depicting the core pillars required for self-driving-laboratories. For more information see: 'Materials Informatics 2020-2030'. Image sources: Collaborative Drug Discovery, Cambridge Crystallographic Data Center, and SAI-TECH
The role of laboratory informatics and robotics can take numerous forms. This can include well known high-throughput experimentation through to full digital platforms and integrated sensors to monitor experiments. Stepping away from the potential human-less end goal, these developments are having an immediate impact on the reproducibility, capacity to internally share, safety, and rate of generating experimental data.
Materials informatics (or cheminformatics or bioinformatics as appropriate) plays a key role in each stage of the experimental cycle. Previous articles have gone into detail of how this can benefit, but from candidate screening and retrosynthetic predictions through to structure-property relations and further analysis, the impact this can have on a closed-loop laboratory process is evident.
Work from the Harvard University, University of Toronto, and University of Glasgow are some of the key institutes in this field, with Kebotix and DeepMatter Group being exciting spinouts commercialising these developments.
There are already notable examples of early versions of this final goal. One early study was demonstrated by the US Air Force Research Laboratory in collaboration with Lockheed Martin. By combining high-throughput CVD synthesis of SWCNTs with AI-led techniques they created an Autonomous Research System (ARES), they demonstrated that the system could learn to optimise the growth of nanotubes by controlling various the experimental parameters.
There has been a recent acceleration in demonstrations, in 2020 the North Carolina State University and the University at Buffalo showed a proof-of-concept in which an appropriate quantum dot could be identified and produced in less than 15 minutes for any colour. Similarly, work from the University of Glasgow explored coordination chemistry through the discovery of new supramolecular complexes with an 'autonomous chemical robot'.
IDTechEx has released a detailed market report on materials informatics, through extensive primary interviews and leveraging deep knowledge in relevant vertical sectors, this report provides the reader with a detailed understanding of the commercial potential in this field.