Catalog will enable Alysophil to offer novel solutions to meet needs of its partners in defense and space sectors, as well as cosmetics, flavors and fragrances
Alysophil SAS, developers of a new industrial approach based on continuous flow chemistry, today announces the creation of a library of one million molecules that have not yet been synthesized. The library was built using several deep neural networks trained since 2017. Its creation is part of an approach combining deep learning technologies with deep tech industrial activity.
Alysophil aims to develop a new industrial chemistry based on innovative processes together with its artificial intelligence tools (ALChemAI). This project, which was end to end led by Alysophil’s R&D team to support its industrial teams, shows that it is possible to create molecules in silico for a specific market, using connected AI algorithms. This database will allow Alysophil to meet the challenges of complex customer and stakeholder requirements and produce the relevant molecules on a short-cycle basis, using new syntheses made possible by flow chemistry.
In addition to their molecular structures, the database includes the physical and chemical properties of the molecules, along with any societal or economic data related to human perception (e.g. odor, toxicology, potential applications, economic sector and synthesizability). It will be regularly updated, whenever an underlying algorithm is improved and as new algorithms and data become available.
“This marks an important milestone for our company,” said Philippe Robin, CEO and co- founder of Alysophil. “With this catalog we will be able to promptly offer novel solutions to meet our partners’ needs, while setting out manufacturing processes with a reduced environmental impact, by using flow chemistry manufacturing (biosourced raw materials, reduced use of solvents and energy, reduced waste and the ability to manufacture at various scales). This work perfectly represents our philosophy in action: flow chemistry empowered by artificial intelligence.”
This repository was built by combining a molecular generator that uses latent space and a set of neural networks trained on criteria such as toxicology, organoleptic properties, economic sectors and markets, with some models also linked to genetic algorithms. As the technologies in use are constantly evolving, the algorithms and results are recorded in the blockchain with a certified date as proof of precedence.
“Our work meets the challenges in the rapid development of multifunctional molecules for markets that demand significant innovation, such as the defense and aerospace sector and the cosmetics, flavors and fragrances sector,” said Luc Brunet, co-founder in charge of R&D and AI. “Our partners tend to be global companies, primarily based in Europe and India. Our aim is to play an active role in transforming the industry in the OECD economies – in the United States in particular – and in fields such as electronics and other users of high value molecules, where our technologies could lead to breakthrough innovations.”
This development, announced by Alysophil, forms part of its future-oriented approach in smart manufacturing. It demonstrates the relevance of combining deep learning and deep tech capabilities. Development processes are greatly reduced thanks to these innovative tools and approaches, stemming from applied research. In its 2019 report, Accenture said that “artificial intelligence and blockchain have the potential to transform both the chemical industry and the industry as a whole.”