Advances in deep learning and artificial intelligence have enabled the de novo design of large numbers of novel small molecules, fit-for-purpose, based on any number of evaluation metrics. Huge numbers of novel molecules with a high probability of attaining a specific set of goals can now be proposed by algorithms trained on the appropriate data sets. Synthetic accessibility to the designed molecules, however, remains a practical limitation for these generative approaches. Generally, this problem is addressed by expert chemists conducting retrosynthetic analysis on a one-by-one basis. Over the last few years, computer-aided synthetic planning (CASP) has evolved, driven by progress in data-driven machine learning (ML) and artificial intelligence (AI) approaches. Iktos, a chemistry-focused company with expertise in ML and AI has developed a CASP platform called Spaya. This template-based AI retrosynthesis software focuses on prioritizing synthetic routes for individual input molecules. For large data sets, the assessment of synthetic accessibility requires a high-throughput framework. To do so, Iktos has developed an API running on Spaya’s algorithmic engine for library scoring purposes. Integration of the Spaya API into the generative process leads to the generation of synthesizable molecules without penalizing the other objectives or constraints of the molecular generator. Integration of synthetic accessibility is an enabling step to be able to take advantage of the full potential of generative models in real life, i.e.: obtaining synthesizable optimized molecules.

Spaya and the spaya API: AI enhanced retrosynthesis planning

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.