Chemical Engineering and Industrial Chemistry

Guidelines for Machine Learning Yield Prediction from Small-Size Literature Dataset



Synthetic yield prediction using machine learning is intensively studied. Previous work focused on two categories of datasets: High-Throughput Experimentation data, as an ideal case study and datasets extracted from proprietary databases, which are known to have a strong reporting bias towards high yields. However, predicting yields using published reaction data remains elusive. To fill the gap, we built a dataset on nickel-catalyzed cross-couplings extracted from organic reaction publications, including scope and optimization information. We demonstrate the importance of including optimization data as a source of failed experiments and emphasize how publication constraints shape the exploration of the chemical space by the synthetic community. While machine learning models still fail to perform out-of-sample predictions, this work shows that adding chemical knowledge enables fair predictions in a low-data regime. Eventually, we hope that this unique public database will foster further improvements of machine learning methods for reaction yield prediction in a more realistic context.

Version notes

The manuscript was revised in order to make the outcomes more intelligible for the synthetic chemist community. The dataset was analyzed in terms of coupling partner, substrates and ligands categories. Reactions features responsible for the model decisions where also highlighted and discussed.


Thumbnail image of ChemRxiv2.pdf

Supplementary material

Thumbnail image of SI.pdf
Supplementary Informations
Details on the code and the methods used to train the model and featurize the data. Additional information supporting the main manuscript.

Supplementary weblinks

NiCOlit code and data
The NiCOlit dataset is available. The code used to generate the results is available.