Computer Vision for Analytical Chemistry (CVAC): Scalable Productivity for Chemical Manufacturing

KEYWORDS: chemistry, catalysis, image processing, manufacturing, productivity, EPSRC.

The digital eyes of cameras paint the colourful worlds we can and cannot see by numbers. These numbers have the power to help us make life-changing medicines on timescales that, at the present time, we cannot imagine.

This research and leadership programme focusses on developing the analytical power of digital cameras to improve the productivity and safety of chemical manufacturing.

The Pharmaceutical sector is the UK's second largest in terms of income, but it is an extremely costly business to run. This cost burden hits at the heart of one of the UK's biggest challenges: our lack of productive output versus hours worked compared with other nations. To this point, 'Big Pharma' stands to realise a >£1bn reduction in R&D cost by 2030, but only if the efficiency with which it can discover new medicines can be improved by a third beyond the current state of the art. How can we make new medicines more productively? Fast adoption of digital technologies is vital. As linked to the core of the proposed research, digital technology adoption should include the amazing ability of cameras to tell the story of the world, not in words, but in useful numbers.

In Big Pharma, to understand whether or not the chemical process of making a medicine is safe to use on the manufacturing scale, we need to be able to analyse the chemical process in real time. The better we analyse a process on the small scale, the better its chances of being used productively to make medicines on the large scale. However, many useful reactions are never applied in industry because they do not meet the strict criteria for safe application on the manufacturing scale. This is an unsolved problem, and no current chemical monitoring technologies can seamlessly analyse chemical processes on small lab scale, large plant scale, and in dangerous environments. If such a monitoring technology were available, it has the potential to lead to an up-to 9:1 return on investment, moving us closer to the ultimate goal of improving research productivity by a full third.

Computer Vision is the science of digitally quantifying real-world objects using cameras. It is a vibrant area of research with a rich history in astronomy, land surveys, autonomous systems, food safety, defence and security, and art forensics, among other areas. Whilst 'photo-style' camera analysis has been used over the past decade, new and unique methods of using real-time camera-based chemical monitoring is still hugely underdeveloped across chemical manufacturing, despite the wealth of emerging knowledge from seemingly unrelated scientific disciplines. The untapped technology of camera-enabled reaction monitoring thus holds remarkable fundamental research potential. A new research programme in this area would contribute strongly to UK chemical manufacturing, realising significant and digitally-adoptive increases in productivity 2-3 years ahead of current 2030 targets.

This ambitious research programme will deliver a world-leading suite of new camera-enabled analytics for understanding a wide range of valuable chemical processes to make them safer and more productive on scale. The research leader has an emerging track record which has already directed step-changes in homogeneous catalyst design, reaction kinetics platforms, safety software systems, and industrial technology translation. Bordering chemistry and computer Science, this programme will deliver research excellence in video analysis methods for visible and invisible chemical processes, across all scales of chemical development, and in a wide range of chemistries beyond the core focus of improving productivity in Pharmaceutical development.

The economic and societal impact of this research in the longer-term will contribute to a 30% increase in Pharmaceutical R&D productivity, £1.3bn savings, and 1,555 fewer accidents associated with Pharmaceutical Manufacturing per year. The impact activities below aim to fully exploit the ability of the CVAC research programme to contribute to these productivity-focussed targets, serving stakeholders at the school-, academic-, industrial, and policy-making levels.

KEY BENEFICIARIES: As outlined in significant detail in the Beneficiaries section, it is envisaged that the proposed programme will serve:
- academic collaborators: colleagues working in electrochemistry, biochemistry, and materials design will benefit from access to new methods of monitoring changes in their systems that were previously difficult or impossible to measure.
- industrial process chemists: gaining unmet access to new high throughput reaction discovery methods that are, unprecedentedly, also applicable to scaled-up processes
- software engineering job seekers: ambitious computer scientists will gain opportunities to design and maintain commercial software emerging from the research.
- school children and coding enthusiasts: outputs from applying computer science to chemistry will be redressed in fun, accessible classes for the next generation of scientific talent.
- popular science content consumers: public lectures and massive open online courses will be created to engage a worldwide audience on the importance of cameras in chemical science.
- international chemistry policy makers: case studies from the research will inform new analytical chemical nomenclature and provide evidence for accelerating digital technology adoption.

TIMING: Chemical Manufacturing is increasing digitalisation, investing $3.2billion by 2022. Emerging 5G internet and increasing accessibility of camera technology are providing infrastructure to deliver connected, data-rich services to Chemical and related sectors. Development of computer vision for analytical chemistry (CVAC) methods for understanding high-value chemical reactions is thus extremely well-timed.

WIDER INFLUENCE: Adoption of digital innovations can reduce R&D spend by £1.3bn (~25%). Productivity improvements of 30% are targeted by 2030. Even if this programme positively influences only 1 of 5 project partners, from over 3,000 companies operating in the UK, a contribution to R&D costing-saving of over £43m could be realised. Non-invasive CVAC kinetic methods developed herein will improve high-throughput experimentation and support more efficient large-scale manufacturing. This helps improve quality control and process safety measures. Overall, this programme will benefit practitioners in catalysis research, materials development, and drug discovery. This is evidenced by the broad academic and industrial collaborations to develop computer vision kinetics beyond the core programme.

MAXIMISING IMPACT: The core research milestones (Y1-4, Work Plan) show escalating dissemination activity for the envisaged reach of the research. After review, Y5-7 will broaden research capabilities towards understanding metal-based carbonyl releasing therapeutics, a vibrant area of drug discovery and natural progression from initial research developments in metal catalysis. Later years will involve additional secondments to each partner company to streamline deployment of the new computer vision kinetics capabilities in-house for company-specific research projects. In relation, technology transfer via software licensing will escalate after the 4-year review. Finally, teaching material development will tackle the highlighted issue of a skills shortage in modern chemists via an increase in chemistry-specific computer programming abilities. Maximising research exposure will also demand a high-quality group website, strategic social media presence, YouTube research video summaries, podcast appearances, and press releases.