(499d) Combined Synthesis Planning for Synthetic Drugs on the WHO Essential Medicines List

Authors: 
Gao, H., Massachusetts Institute of Technology
Coley, C. W., Massachusetts Institute of Technology
Li, L., Harvard T.H. Chan School of Public Health
Qian, Y., MIT
Green, W. H., Massachusetts Institute of Technology
Jensen, K. F., Massachusetts Institute of Technology
The WHO Model List of Essential Medicines (EML) was created in 1977. It contains medicines that are needed to satisfy the priority healthcare demand and should be in sufficient amount in any time[1]. However, up to today, access to essential medicines is still problematic for about one-third of the world’s population. Prices for some medicines make them unaffordable in less developed areas[2], and drug shortages have been long and widespread challenges[3]. While this is a complicated problem and no single solution can suffice, it is desirable to develop strategies to lower the manufacturing costs of the medicines and simplifying the supply chain for the essential medicines for quicker response to drug shortages.

In this work, we take advantage of the recent development in artificial intelligence applied in synthetic planning to explore solutions to the above challenge. Using a state-of-the-art synthesis planning tool, we try to minimize the chemical inventory needed to access the synthetic drug compounds in the WHO EML. The problem is solved in two stages. In the first stage, retrosynthesis analysis is performed for all the target molecules. Machine learning models are used to help find synthetic pathways[4], propose reaction conditions[5], and evaluate the feasibility of the suggested reactions[6]. In the second stage, a multi-objective mixed-integer programming problem is solved[7] to minimize the chemicals used in all the syntheses while maximizing the likelihood of success for all the pathways. Results show that by planning the syntheses in a combined manner we can reduce the number of chemicals by 30% compared to planning the syntheses separately, without significantly sacrificing the likelihood of success of the syntheses. In the meantime, this work demonstrates the potential benefit of simultaneously planning the syntheses for multiple targets, which can also be applied to other multitarget lists, e.g., a compound library during drug discovery.

Reference

  1. The World Heath Organization. WHO Model Lists of Essential Medicines Available online: https://www.who.int/medicines/publications/essentialmedicines/en/.
  2. Hill, A.M.; Barber, M.J.; Gotham, D. Estimated costs of production and potential prices for the WHO Essential Medicines List. BMJ Glob. Heal. 2018, 3, e000571,
  3. Gray, A.; Manasse, H.R. World Health Assembly». Bull. World Health Organ. 2012, 90, 158–158A.
  4. Segler, M.H.S.; Waller, M.P. Neural-Symbolic Machine Learning for Retrosynthesis and Reaction Prediction. Chem. - A Eur. J. 2017, 23, 5966–5971
  5. Gao, H.; Struble, T.J..; Coley, C.W..; Wang, Y.; Green, W.H.; Jensen, K.F. Using Machine Learning to Predict Suitable Conditions for Organic Reactions. ACS Cent. Sci. 2018, 4, 1465–1476
  6. Coley, C.W.; Jin, W.; Rogers, L.; Jamison, T.F.; Jaakkola, T.S.; Green, W.H.; Barzilay, R.; Jensen, K.F. A graph-convolutional neural network model for the prediction of chemical reactivity. Chem. Sci. 2019, 10, 370–377.
  7. IBM Co. IBM ILOG CPLEX. 2018