A novel deep learning-based prediction modeling approach using molecular image in the activation of progesterone receptor

Abstract

In silico analysis have been expected as an alternative for animal testing of toxicity. Recently, deep learning (DL) has a focus on image classification. Thus, a novel DL-based quantitative structure-activity relationship method, called DeepSnap-DL that applied molecular image into DL, has been developed in our laboratory. This study revealed that DeepSnap-DL shows high-prediction performance in the activation of the progesterone receptor that could mediate key initiating events on adverse outcome pathway, and outperforms conventional machine learning technics.