(584d) Identifying Potential One-Pot Synthesis from Computer-Planned Synthetic Routes

Recent years have seen rapid advancement on the development of computer-aided synthesis planning (CASP) tools. Using expert-crafted rules¹ or machine learning from chemical databases^2,3, these tools have become so powerful as to attract the attention of organic chemists, especially in the pharmaceutical industry. While good at idea generation, these CASP tools easily find numerous pathways, and prioritizing them is not a trivial task. Among the many practical considerations that a chemist would use to evaluate the synthetic routes, the possibility for one-pot synthesis development is an important aspect. Ideally in a one-pot synthesis, all reaction steps are carried out consecutively without needing to isolate intermediates. It has the benefit of minimize material loss, risk of leakage, and ease of operability.⁴ To successfully develop a one-pot synthesis, it is important that the chemical species do not interfere with the subsequent reactions.

In practice, the development of one-pot synthesis involves different considerations, and in this work we explore one aspect – the similarity of reaction conditions between consecutive reaction steps. The assumption is that similar conditions will be more likely compatible with each other and not causing side reactions. While this alone does not guarantee the success of a one-pot synthesis, we aim to use this as a prioritization method at the synthesis planning stage. We explore two different approaches to identify one-pot synthesis from computer-planned synthetic routes, focusing on the similarity of reaction conditions – implicit and explicit. The implicit approach uses embedding of reaction conditions from machine learning models⁵ to quantify similarity of reaction conditions. The explicit approach uses models to predict a number of sets of reaction conditions for each reaction and compare them directly. The average similarity of reaction conditions between consecutive reaction steps is used as a means of quantifying the suitability for one-pot synthesis development. A number of computational case studies are performed to qualitatively demonstrate the effect of these approaches.

Reference

1 S. Szymkuć, E. P. Gajewska, T. Klucznik, K. Molga, P. Dittwald, M. Startek, M. Bajczyk and B. A. Grzybowski, Angew. Chemie Int. Ed., 2016, 55, 5904–5937.

2 C. W. Coley, D. A. Thomas, J. A. M. Lummiss, J. N. Jaworski, C. P. Breen, V. Schultz, T. Hart, J. S. Fishman, L. Rogers, H. Gao, R. W. Hicklin, P. P. Plehiers, J. Byington, J. S. Piotti, W. H. Green, A. J. Hart, T. F. Jamison and K. F. Jensen, Science (80-. )., 2019, 365, eaax1566.

3 M. H. S. Segler, M. Preuss and M. P. Waller, Nature, 2018, 555, 604.

4 Y. Hayashi, Chem. Sci., 2016, 7, 866–880.

5 H. Gao, T. J. . Struble, C. W. . Coley, Y. Wang, W. H. Green and K. F. Jensen, ACS Cent. Sci., 2018, 4, 1465–1476.