ARTIFICIAL INTELLIGENCE MODEL FOR PREDICTING INDICATIONS FOR TEST SUBSTANCES IN HUMANS

摘要

The present invention addresses the problem of, even in the case where a test substance has an efficacy not known for existing substances used in the acquisition of training data, predicting said efficacy.　The present invention uses an artificial intelligence model trained by a training method comprising: training the artificial intelligence model by associating a first training data set, a second training data set, and a third training data set, and inputting the same into the artificial intelligence model. The first training data set is a set of data in which a set of data indicating the behavior of biomarkers in one or a plurality of different organs collected from non-human animals that were individually administered a plurality of predetermined existing substances that have known indications in humans is linked to labels indicating the respective names of the plurality of predetermined existing substances that were administered. The second training data set is a set of data in which the labels indicating the respective names of the plurality of predetermined existing substances are linked to labels indicating the indications reported for the respective plurality of predetermined existing substances. The third training data set is a set of data in which the labels indicating the indications reported for the respective plurality of predetermined existing substances are linked to information pertaining to reported adverse events corresponding to the respective indications. The artificial intelligence model is used for predicting indications for test substances in humans.