University of Tokyo chemists extract reaction-step rates from the yield data they gather while optimizing

Kinetics has usually been a separate job. To learn how fast each step of a reaction runs, chemists set up dedicated experiments that follow the reaction over time ^[4]. Optimization is its own campaign, varying concentration, temperature or the timing of additions to find the conditions that give the best yield ^[15]. "For years, chemists have treated these as two different objectives. We optimize reactions to achieve the highest yield, and if we want to understand the mechanism, we perform another series of experiments that follows the reaction over time," said Hiroyuki Isobe, a professor in the University of Tokyo's Department of Chemistry ^[5][13].

In CYAN, the chemist supplies the chemistry. The model fills out the yield data, but the rate equations, and with them the proposed reaction pathway, come from the researchers ^[6]. "In this way, a single set of experiments can be used for both reaction optimization and kinetic analysis," Isobe said ^[6]. He called the approach "a two-way collaboration between machine learning and organic chemists, in which we contribute the chemical hypotheses and experimental data, while machine learning contributes a wealth of augmented yield data" ^[14]. I think this split suits organic chemistry, where the pathway is the thing people want to test. It also means the rate constants are only as sound as the rate equation someone chose to fit ^[6].

The project started from the opposite problem. The group's earlier machine-learning tool "was very good at finding better reaction conditions, but it behaved like a black box, telling us what worked without telling us why," Isobe said ^[9].

Nickel's role in the ring synthesis is the only test the release describes, and the counterintuitive part was the slowdown ^[10][11]. "The nickel result surprised us because it slowed part of the reaction, and that sounds like it should make synthesis worse, not better. A useful way to imagine it is water flowing through a river that splits into two channels. If one channel becomes narrower, more water naturally follows the other route," said Isobe ^[12].

On this evidence, CYAN is a credible way to get first estimates of step rates from data a lab already holds ^[1]. Standing in for time-course experiments is a larger claim. The university's account describes a single reaction. It does not say how many optimization runs fed the analysis, or whether the fitted constants were compared with a conventional kinetic measurement of the same system ^[10].