(175a) Extractionscore: A Quantitative Framework for Evaluating Synthetic Routes on Predicted Liquid-Liquid Extraction Performance

A multitude of metrics exist to assign scores to synthetic routes within computer-aided synthesis planning (CASP) tools. A quantitative scoring method is necessary to identify the most promising synthetic approaches to a molecule. Popular metrics include reagent cost, [1] quantifications of synthetic complexity, [2] and confidence in predicted reaction outcomes and diversity of reactions employed. [3] However, relatively little attention in the CASP field has been paid to the impact of side products on separation sequences, in part due to a lack of experimental data.

To address this issue, we have developed a method of quantifying liquid-liquid extraction performance for synthetic routes, which we call ExtractionScore. ExtractionScore uses existing reaction outcome predictors to predict side products in a reaction mixture, and chemical property prediction to estimate partition coefficients. Separation factors for impurities in the reaction mixture are then calculated and used to assign a numeric score for the reaction. The ExtractionScore for a synthetic route is calculated from the scores of each reaction in the route.

We have evaluated ExtractionScore by comparing the industrially practiced synthesis routes to 200 active pharmaceutical ingredients and precursors thereof with routes suggested by the IBM RXN retrosynthesis planner. In 72% of cases, the industrially practiced route scores higher on ExtractionScore than the IBM RXN suggested routes, indicating ExtractionScore’s potential usefulness as a quantification of industrially relevant features of a synthesis route.

1. Szymkuć, ; Gajewska, E. P.; Klucznik, T.; Molga, K.; Dittwald, P.; Startek, M.; Bajczyk, M.; Grzybowski, B. A. Computer-assisted synthetic planning: The end of the beginning. Angewandte Chemie International Edition 2016, 55, 5904–5937.
2. Coley, C. W.; Rogers, L.; Green, W. H.; Jensen, K. F. SCScore: Synthetic complexity learned from a reaction corpus. Journal of Chemical Information and Modeling 2018, 58, 252–261.
3. Schwaller, ; Petraglia, R.; Zullo, V.; Nair, V. H.; Haeuselmann, R. A.; Pisoni, R.; Bekas, C.; Iuliano, A.; Laino, T. Predicting retrosynthetic pathways using transformer-based models and a hyper-graph exploration strategy. Chemical Science 2020, 11, 3316–3325.