Published
24 May 2024
The Chemputer®: how far are we from it?
The ultimate manifestation of the Chemputer® (chemical computer) is a system able to synthesise a molecule in a completely automatic way (without any human intervention). Ask the Chemputer® to produce a sample of a given molecule, come back a few days later, and get a sample of the molecule you requested, produced in total autonomy by the Chemputer® with high yield and purity. The Chemputer® will enable fast discovery of new molecules and development of cost-effective and sustainable processes to produce these molecules. The economic, societal and environmental impact of such a technology is obviously immense.
The Chemputer® will be composed of several parts:
- Data, to power predictive and prescriptive Machine Learning algorithms
- Software, (including machine learning algorithms), that designs experiments, controls the hardware to run the experiments and processes the data produced by the experiments
- Hardware, to perform chemistry experiments (by executing machine-readable code) and measure their outcome
As of today, deepmatter® already has a strong technology base, protected by granted patents, that is precursor for full Chemputation. That technology comprises:
- Synthetic accessibility scoring, allowing to estimate how easy/difficult a molecule will be to synthesise
- Synthetic route design (retrosynthesis), suggesting a several putative synthetic routes for a given molecule, with the ability to filter these routes based on criteria including the price of chemicals or their sustainability
- Database searches, to identify previous instances of a given reaction
- Bayesian optimisation to find optimal reaction parameters within a design space chosen by the user, either in open-loop (with intervention of a human to run the suggested experiments) or in closed-loop (fully automated)
- Rich data collection (including recipes, time-course data collected by sensors and events), allowing to build more accurate machine learning models and better data analysis
- Algorithms that generate machine-readable code to enable robotic control
- A proprietary data set of several millions of chemical reactions, powering deepmatter®’s machine learning algorithms
The fact that we are able to automatically perform retrosynthetic analysis and that we have already performed closed-loop optimisation of chemical reactions, in batch and continuous reactors, is a demonstration that the Chemputer® will become a reality in the coming years.
However, we have not yet achieved true and complete Chemputation. We are working hard on it and have several important milestones ahead of us:
- At the end of each experiment, feed back the data to the reaction database so that it gets used during further retrosynthetic analyses, thus augmenting them
- Wider automation capabilities, including lab orchestration and integration of more robotic hardware
- Integration of more analytical equipment (online and offline)
- Integration with common Electronic Lan Notesbooks (ELNs)
- Inventory management
- Account for sustainability for predictions/suggestions
- And many more..
Stay tuned for news about our progress towards the Chemputer®. If you want to be part of that journey and join our industrial and academic partners, please contact our team of experts.
“CHEMPUTER” is a registered trademark of deepmatter®