Reinvestigation of a robotically revealed reaction

Autonomous Discovery in the Chemical Sciences Part I: Progress

“…training this binary classifier are misaligned:T he algorithmic exploration of the search space of possible reactions does not direct experiments towards those likely to lead to novel reactions;the reactions claimed to have been discovered were identified only through (error-prone [310] )m anual analysis of product mixtures.M oreover,a ll 969 reaction combinations could have been performed in am iniaturized well-plate format while using less time and materials (see item (4) in Section 3.1) and the one-hot encoding of substrates precludes prediction of reactivity for unseen substrates.A ne arlier version of this platform was used by the same group to explore the 64 possible pathways defined by at hree-step synthesis with one of four reagents added at each step. [311] The "most reactive" pathway was found through astepwise greedy search by identifying the reagent whose addition led to the largest change in the ATIR spectrum of the product mixture.…”

“…The physical apparatus is reminiscent of MEDLEY—an automated reaction system employing computer‐controlled pumps connected to a round‐bottom flask [309] . However, the goals of reaction discovery and training this binary classifier are misaligned: The algorithmic exploration of the search space of possible reactions does not direct experiments towards those likely to lead to novel reactions; the reactions claimed to have been discovered were identified only through (error‐prone [310] ) manual analysis of product mixtures. Moreover, all 969 reaction combinations could have been performed in a miniaturized well‐plate format while using less time and materials (see item (4) in Section 3.1) and the one‐hot encoding of substrates precludes prediction of reactivity for unseen substrates.…”

Autonome Entdeckung in den chemischen Wissenschaften, Teil I: Fortschritt

“…Die Plattform selbst erinnert an MEDLEY – einen Reaktionsautomat mit computergesteuerten Pumpen, die an einen Rundkolben angeschlossen sind [309] . Allerdings passt das Training dieses binären Klassifikators nicht zum Ziel der Entdeckung von Reaktionen: Die algorithmische Untersuchung des Suchraumes möglicher Reaktionen lenkt die Experimente nicht zu denen, die wahrscheinlich zu neuartigen Reaktionen führen; die mutmaßlich entdeckten Reaktionen wurden nur durch (fehleranfällige [310] ) manuelle Analyse von Produktmischungen identifiziert. Darüber hinaus wären alle 969 Reaktionskombinationen im miniaturisierten Mikrotiterplattenformat weniger zeit‐ und materialaufwändig gewesen (vgl.…”

Real-time prediction of 1H and 13C chemical shifts with DFT accuracy using a 3D graph neural network

“…Comparisons of experimental isotropic chemical shis (i.e., those measured for solution samples) with computationally predicted values have been applied, sometimes including scalar coupling constants, to various problems in structure elucidation: the assignment of relative stereochemistry in exible organic molecules as pioneered by Bagno and Bifulco, 1-3 complex natural product structure elucidation and reassignment, [4][5][6] identication of the side product(s) in synthetic reactions, 7,8 deducing the macromolecular conformation adopted by cyclic peptides, 9 and in correcting literature misassignments. 10 The growing importance of computational chemical shi prediction, particularly of 13 C and 1 H nuclei, in natural product, mechanistic and synthetic organic chemistry is the subject of an authoritative review by Tantillo and co-workers. 11 To serve as a useful tool for structure elucidation, prediction errors in computed chemical shis must be smaller than the experimental variations between different candidate structures.…”