SyntheticGestalt, a research and development company specializing in the application of AI to the life sciences domain, and Enamine, the world's leading provider of chemical building blocks, screening compounds, and integrated drug discovery services, have announced the start of a joint effort to create a suite of AI models that will enable the generation of synthetically accessible biologically active compounds with optimized physicochemical and ADME/ Tox properties. The models will be applicable to the compound discovery initiatives of SyntheticGestalt, as well as its service for both academic users and pharmaceutical companies.

Enamine will provide access to its largest enumerated database of make-on-demand compounds, Enamine REAL database, which has 38 billion molecules in its current edition. SyntheticGestalt will add the REAL database to its Drug Discovery Service, which uses proprietary AI models to provide predictions on physicochemical and ADME/ Tox properties of compounds. For compounds with issues, the service proposes improved alternative compounds instantaneously.

Enamine will synthesize the selected compounds within just 3-4 weeks and provide quality pharmacological in vitro profiling data through the in-house tests to streamline and shorten the discovery cycle.

Furthermore, SyntheticGestalt will enhance its pre-trained AI model using the data provided by Enamine. It is expected to become the largest pre-trained model in the world based on the 3D structures of the compounds, to improve the predictive accuracy of SyntheticGestalt's machine learning models. The resulting models will be offered on a joint research basis to certain interested parties. The pre-trained AI model and its performance will be presented at NVIDIA's annual event, NVIDIA GTC Japan AI Day, in March of 2024.

Iaroslava Kos, Director of Business Development at Enamine commented: "The promise provided by AI/ML powered computational designs in the discovery of new drugs cannot be underestimated. Finding new active compounds by synthesis of just a handful of novel compounds looks fantastic. We are pleased to enter collaboration with SyntheticGestalt bringing to the table the talent and expertise of our scientists to realize mutual goals. We look forward to joining efforts towards the long-desired therapeutics."

Koki Shimada, CEO at SyntheticGestalt commented: "The use of machine learning in drug discovery research has long been plagued by the problem of high performance on training data but low performance in actual use. To solve this problem, it is necessary to develop pre-training models using data that will be used in real-world applications. The Enamine REAL database is the perfect match for our initiative as the most trusted and the largest make-on-demand set on the market. We believe that the ultra-large pre-trained model we are developing will enable a cosmic leap in AI drug discovery, just as the large-scale pre-training made a revolution in Large Language Models (LLMs)."