DESIGN AND OPTIMIZATION OF BIOCATALYTIC CONTINUOUS MANUFACTURING OF DIABETES DRUG THROUGH MACHINE LEARNING

Abstract

This work proposes a green-engineering framework of biocatalytic continuous manufacturing of sitagliptin, the active pharmaceutical ingredient of the leading dipeptidyl peptidase-4 inhibitor antidiabetic drug. An end-to-end continuous manufacturing process is designed and the reaction kinetics of the biocatalytic reaction is determined according to the published data. A comprehensive techno-economic analysis is performed, validating the economic feasibility of this process with a net present value of $150 million over a 20-year time period. The results of life cycle assessment reveal that 547.8 kg of carbon dioxide and 1.78 kg of sulfur dioxide per kg sitagliptin monophosphate produced would be caused. Among all the chemicals, trifluoroacetic anhydride is identified to cause over 30% impact in most impact categories. To design a greener process, out of two million candidate alternatives, 30 most similar chemicals are identified using molecular structure-based methods, and their life-cycle impact is predicted using optimized deep neural network and ensemble machine learning. The results demonstrate that the chemical 1,1,1,5,5,5-Hexafluoro-4-methylpent-3-en-2-one is one of the potential greener substitutes, and neural network with exponential linear unit and simpler structure, such as 2 hidden layers and 25 neurons, shows better prediction performance.