Evolving and Nano Data Enabled Machine Intelligence for Chemical Reaction Optimization

02 November 2018, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

Optimizing reaction conditions is an essential routine in synthetic chemistry. However, selecting appropriate experiments remains tightly connected to expert chemistry knowledge. Here, to streamline the reaction yield optimization process and disconnect it from chemical intuition, we developed an adaptive machine intelligence to navigate multidimensional reaction conditions’ spaces. Our approach (LabMate.AI) employs an interpretable algorithm and requires only <0.05% of all search space as input data. LabMate.AI optimizes many reaction parameters simultaneously, and uses minimal computational resources and time. We demonstrate how LabMate.AI can identify optimal conditions for a Ugi and a C–N cross-coupling reaction in a more efficient and faster manner than human experts, while affording reactivity insights. Our approach formalizes chemical intuition, and acquires expert chemistry knowledge autonomously, thereby providing an innovative framework towards informed and automated experiment selection. The results support machine learning for hastening experimental design, democratizing synthetic chemistry, and freeing chemists for non-routine tasks.

Keywords

Active Learning Methodologies
reaction optimization experiment
Design of experiments
Synthetic chemistry

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.