Complex Chemical Systems Platform Exploring Inorganic Intelligence

Lead Research Organisation: University of Glasgow
Department Name: School of Chemistry


Our vision is to establish the new field of inorganic intelligence by defining the key fundamental science problems, and by developing researchers equipped with the right skills to explore this emerging area of science. The Cronin Group has made world-leading contributions to foundational aspects of this research and now we need to explore, unify, and develop some of the central science problems. These include how to explore and control, and understand complex chemical systems using robotics and real-time data. We anticipate that the coordinated development of these four topics will lead into applications as diverse as self-assembly control in nano molecules, chemical synthesis and discovery automation or artificial intelligence (AI) optimisation of reactions and exploration and discovery of new underpinning principles. The new grant will continue to unify and develop synergies already established during the previous Platform, but most importantly will ensure continuity and stability. This will enable the team to evolve from focusing on inorganic systems to the digital control and exploration of complex chemical systems. The new Platform will not only contribute to unify the many strands already existing in the team, but will also allow an extension to new disciplines including robotics, machine learning, and development of synergies across those areas - a combination of topics very rarely merged and hence extremely hard to raise funding using other mechanisms. Thus, the new Platform is essential for continuation and the evolution of the research activity, giving added value in integrating the group, allowing us to be strategic and develop the team into the chosen new areas defining the area of 'inorganic intelligence'. The previous grant was instrumental in letting us extend our critical mass, enhance key existing international collaborations, and support inter-group collaborations in Glasgow, which allowed us to speculate and develop our exploratory work in chemical robotics. In addition, we had the flexibility to support and further consolidate some of the existing team, and to hire in new expertise, as well as restructure the team with help from the EPSRC mentor scheme. We need the new platform to continue our team development and provide stability and flexibility especially important during the next few years. As before, we will aim for our best results to be published in Science and Nature, protect innovations by patent applications, and engage a user group and industrialists as well as other world-leading academics to maximise both the academic and technological impact. This will be achieved by making full use of funding from various sources, aiming at areas that need to be developed using the Platform as a consolidating component. We will also seed 'pump-prime' projects within the Platform, provide bridging funding, and be ready to exploit unexpected and high impact results. The Platform will ensure the group remains at critical mass at a critical time, and at the cutting edge of science in a range of new areas.

Planned Impact

The digitization of chemistry has a trillion-dollar potential as a disruptive technology (remote drug manufacture), to develop a chemical data driven economy (chemistry on the cloud), as well as networked chemical synthesis (distributed chemical infrastructure). This application bridges many of the EPSRC outcome frameworks e.g. resilience and productivity (chemical manufacturing based upon digital code and decentralised), connected (chemical code will be on the cloud and aid collaborative developments in digital space), and healthy (algorithms to power a serendipity engine for the discovery of new drug molecules).

We will disseminate our work as widely as possible through publication in high impact journals. We will aim to publish in open access journals or have our publications on the open-archive within 3 months of publication (aiming for 1 month). We will also build web resources for wide dissemination of data (open data), and have a digital platforms website established to help translate affordable robotic systems for use by chemists between labs. We will also ensure that our open / collaborative agenda will be advertised through the large number of invited talks given by the team both in the UK and abroad, including major national and international conferences. We intend to interact directly with the Glasgow University Impact Agenda (GUIA) (including EPSRC-funded Impact Acceleration Accounts) as well as the Cronin-group Glasgow University Research and Enterprise Officer (Melville Anderson). These existing resources will provide support with strategic management of IP. The University of Glasgow prides itself of being the instigator of an innovative, award-winning model for the management of IP known as Easy-Access IP. This represents the University's commitment to maximize the impact from research by adopting a flexible approach to interactions with industry. The GUIA is a key element that we will exploit to give this work visibility, interact with end-users, and develop a forum of interested parties that will receive information and progress updates about the project as it proceeds. Both Scottish Enterprise and the IP Group (technology investors who have a partnership with Glasgow University) have been interacting with us and have been evaluating the overall portfolio with a view to developing a range of investments in order to develop exploitable technology, know-how, and product innovations.

Also, engagement with the companies will allow us to write a road-map developing the platform beyond inorganics to organic synthesis (championed by GSK) and reactionware (championed by MilliPoreSignma). SpacePharma is interested in developing flight-ready miniature chemical synthesis laboratories for drug synthesis in low earth orbit. We will allow the companies to 'embed' researchers in the team and with the platform for collaborative visits / projects as well as encouraging some of our team to second to industrial sites for training and knowledge exchange.
Description Our interest is in combining machine learning and robotics to enable focused exploration for new coordination complexes, without prior simulation. Autonomously exploring real chemical systems for truly novel molecules requires overcoming key challenges: platforms should be capable of long-term operation, learning programs should be unsupervised (i.e. not require a training dataset), and analysis must be capable of the deconvolution of complex mixtures.

We have developed a chemical robot to address these challenges. The robot is based on a non-deterministic liquid-handling system capable of searching a pre-defined chemical space for novel coordination complexes, using an unsupervised search algorithm to 'rediscover' chemical rules. This has led to the isolation of multiple new complexes and additional in situ observations.
Exploitation Route This system can be used to find new 'rules' in inorganic chemical synthesis which could be applied to a wide range of fields including functional materials, catalysis and optoelectronics.
Sectors Chemicals,Electronics,Pharmaceuticals and Medical Biotechnology

Title The Chemputer 
Description A universal modular robotic synthesiser which can undertake ca. 60% of the batch reactions in the chemical literature. This also includes the XDL language and ontology for translating chemical procedures into universally readable actionable code which can potentially be implemented in any robotic system. 
Type Of Material Improvements to research infrastructure 
Year Produced 2017 
Provided To Others? Yes  
Impact 19 News outlets have reported on this discovery. Plans are underway to setup a spinout and patent aspects of the discovery. 
Title The rapid electrochemical activation of MoTe2 for the hydrogen evolution reaction 
Description The data set includes raw experimental files of electrochemical data, XRD and Raman spectroscopy as well as optimised coordinates for structures from computational studies. A comprehensive description is provided within description file. 
Type Of Material Database/Collection of data 
Year Produced 2019 
Provided To Others? Yes  
Description Integrated Discovery Chemputer Toward Addiction Free Opiates 
Organisation Arizona State University
Department School of Earth and Space Exploration
Country United States 
Sector Academic/University 
PI Contribution The Cronin group are experts in chemical robotics, database development, ELNs, machine learning for chemistry, theory, robotics, in particular the development of closed-loop engines capable of building databases, populating chemistry notebooks, programming chemistry robots, and developing real-time closed loop assays for assessing the real time spectroscopy, structure, and molecular diversity of chemical reactions. We have previously shown that the synthesis of a wide range of organic molecules is possible using our automated Chemputer reactor and that in-line analysis can provide real time data for optimisation algorithms. The main aim of this challenge is to develop and integrate components from multiple platforms into a unified chemical synthesis platform. To do this we have developed Chemical Description Language (XDL) to formalise the way chemists execute and report chemical protocols and synthesis procedures. This language allows the modular use of a wide range of hardware to carry out bespoke chemical synthesis.
Collaborator Contribution The team at ASU are experts in theory developing the networks for retrosynthetic analysis, exploring chemical space, and developing information measures to target novel opiate targets in chemical space using a network theory approach. We will use the Chemputer system and XDL language to search for novel molecules with target properties. To quantify the similarity of the search results to the intended goal we require a fitness function. This function is a mathematical description of the distance from the desired properties. The fitness function could, for example, compare two absorption spectra. We will design sensors so that the directly update the fitness function. This means that new discoverable molecules will be used to update the database and that potential new virtual libraries of accessible molecules will be fed into the database in real time.
Impact Proc. Natl. Acad. Sci. USA, 2019, 116, 5387-5392; Astrobiology, 2018, 18, 779-824
Start Year 2018