Abstract
Organic synthesis plays a vital role in optimizing existing drugs and innovating new drugs. As a significant and
challenging research frontier in the field of organic synthesis, cross-coupling reactions have also attracted
considerable attention. In the past few years, machine learning has realized great potential in predicting the
performance of cross-coupling reactions. However, most of the existing machine learning predictions are based on the
two-dimensional feature information of the cross-coupling reactions. In order to obtain the coupling reaction feature in
a multifaceted way, we exploit the three-dimensional features of the molecules based on the molecular stick-and-ball
model and the persistent homology analysis of topological data, respectively. On this basis, a weighted light
convolutional neural network with multi-scale subtraction (OS-MSW) is proposed to extract the deep abstract features of
the input data, and the extracted abstract features are applied to LightGBM for yield prediction, thus constructing a
highly efficient prediction system OS-MSWGBM. In addition, the interpretability of the OS-MSW model is analyzed in this
paper. The experiments demonstrate that OS-MSWGBM exhibits higher efficiency and more accurate prediction results, as
well as notably stable prediction performance, which can provide reliable decision-making information for experimental
personnel or organizations.
