FUNAMBULIST: AI to improve chemical synthesis
The optimisation of chemical synthesis processes has historically been one of the great challenges of the chemical industry, seeking maximum productivity, minimum costs and, in the current context, maximum sustainability.
The challenge is particularly demanding in high value-added industries such as cosmetics or pharmaceuticals, where innovation involves working with complex reactions, expensive materials and long development times. In such environments, a significant part of the work occurs before there is an 'industrialisable', stable and optimised process.
Identifying a synthesis route, testing it and turning it into a reliable method often requires numerous rounds of trial and error. Each iteration provides information but also consumes resources.
The frontier between laboratory and industrial production
Let's consider a simple case: we want to synthesise a new molecule with very specific properties and characteristics. Based on scientific literature and previous experience, the research team identifies a preliminary route and selects reagents, catalysts, solvents and guideline conditions (temperature, pressure, time) to obtain certain amounts of the desired compound.
The first problem arises when testing this recipe in the laboratory: the yield is tremendously low, with very low production of the target molecule, unconverted reagents and the generation of unwanted by-products.
From this, reasonable hypotheses emerge: perhaps the mixture contains enantiomers (same molecules, but with different spatial orientations) in the reactants, a high specificity of the catalyst (based on DNA chains) and an apparent low solubility of the whole system in the solvent.
AI redefines the scientific workflow without altering the fundamentals of the traditional method
The diagnosis points to a clear conclusion: the process is inefficient, far from being stable and far from reaching the necessary conditions for industrial scale-up.
In a traditional approach, this point marks the beginning of a long experimental campaign: formulation of new hypotheses, successive trials, analysis of hundreds if not thousands of results, and endless iterations until a truly viable and efficient synthesis route is found.
AI as an ally of the scientific method
In this context, Artificial Intelligence brings a differential value, redefining the general workflow, without altering the fundamentals of the scientific method:
- AI does not replace scientific knowledge: it always starts from a hypothesis formulated by experts, based on data, observations and previous experience.
- The hypothesis leads to the design of experiments, in which all key variables of the system are identified and quantified.
- The initial experimental results generate a dataset, which Artificial Intelligence analyses to detect patterns, correlations and relationships that are not immediately obvious.
- Based on this analysis, new experiments can be designed in a more targeted way, aimed at exploiting the most promising combinations with the highest probability of success.
Artificial Intelligence thus becomes a tool for optimising the scientific method, capable of guiding the iterative process with greater precision and speed. It enables the elimination of ineffective combinations and the refinement — or even refutation — of initial hypotheses, leading to highly efficient solutions supported by a solid statistical foundation.
The role of Izertis in FUNAMBULIST
Izertis participates in the European project FUNAMBULIST (FUnctional Nucleic Acids as Versatile SMart BUilding BLocks in Non-ConventIonal SolvenTs), funded by Horizon Europe's EIC Pathfinder programme. The project contributes to the development and testing of artificial intelligence tools aimed at supporting the optimisation of synthesis processes.
FUNAMBULIST relies on the collaboration between different European partners, including the University of the Basque Country (coordinator), the University of Bonn, the University of Strasbourg and the University of Lisbon. The combination of capabilities (academic research, experimentation and technological support) allows the problem to be tackled from various angles.
FUNAMBULIST combines research, experimentation, and technology
The ultimate goal is to move towards a more efficient production of fine chemicals, a segment where a significant part of the cost is concentrated in the separation and purification stages.
Improved synthesis conditions can result in fewer by-products, more stable processes and a reduced burden on those subsequent phases.
In this context, the AI developed by Izertis serves as a practical support tool: it helps interpret experimental results and plan the next steps more effectively, with the aim of accelerating progress towards more robust and sustainable processes.