Machine Learning C–N Couplings: Obstacles for a General-Purpose Reaction Yield Prediction

Abstract

Pd-catalyzed C−N couplings are commonplace in academia and industry. Despite their significance, finding suitable reaction conditions leading to a high yield, for instance, remains a challenging and time-consuming task which usually requires screening over many sets of conditions. To help select promising reaction conditions in the vast space of reagent combinations, machine learning is an emerging technique with a lot of promise. In this work, we assess whether the reaction yield of C−N couplings can be predicted from databases of chemical reactions. We test the generalizability of models both on challenging data splits and on a dedicated experimental test set. We find that, provided the chemical space represented by the training set is not left, the models perform well. However, the applicability domain is quickly left even for simple reactions of the same type, as, for instance, present in our plate test set. The results show that yield prediction for new reactions is possible from the algorithmic side but in practice is hindered by the available data. Most importantly, more data that cover the diversity in reagents are needed for a general-purpose prediction of reaction yields. Our findings also expose a challenge to this field in that it appears to be extremely deceiving to judge models based on literature data with test sets which are split off the same literature data, even when challenging splits are considered.

A New Regularized Spatiotemporal Attention-Based LSTM with Application to Nitrogen Oxides Emission Prediction

“…To verify the effectiveness of the proposed RSTA mechanism, an indicator called the permutation variable importance (PVI) 40 was used to measure the importance of each input variable to the output variable. The PVI of an input variable is obtained by calculating the ratio between the prediction errors before and after the variable data are shuffled; only the data corresponding to the variable are shuffled each time.…”

Machine Learning Strategies for Reaction Development: Toward the Low-Data Limit

“…On the other hand, a similar analysis on Buchwald− Hartwig C−N coupling reactions compiled from the literature suggests the opposite: using the entire data set improved the models. 62 At this point in time, recognizing a priori which approach is best is not a clearly defined task. Alternatively, multiple source data sets could be utilized for transfer learning, resembling the traditional approach where a chemist's source data contains just a handful of research articles.…”

“…This showed that for reactions defined using domain specific knowledge, modest predictivity is plausible using relatively small data sets, in this case about 100 data points. On the other hand, a similar analysis on Buchwald–Hartwig C–N coupling reactions compiled from the literature suggests the opposite: using the entire data set improved the models . At this point in time, recognizing a priori which approach is best is not a clearly defined task.…”

Bayesian Optimization as a Sustainable Strategy for Early-Stage Process Development? A Case Study of Cu-Catalyzed C–N Coupling of Sterically Hindered Pyrazines

“…However, there is not a “general catalytic system” and each reactant combination (aryl iodide/bromide/chloride and amine/amide/N-heterocycle/etc.) seems to require a specific copper source and a dedicated ligand . Furthermore, literature precedents mainly focus on relatively simple substrates, whereas examples of more challenging coupling partners typically found in discovery programs (e.g., sterically hindered, , containing multiple nitrogens which can chelate the metal center and inhibit its catalytic activity), are rarer.…”