Abstract
ICH issued Guideline S5 (R3) for Detection of Reproductive and Developmental Toxicity for Human Pharmaceuticals, which was reported by PMDA on January 29, 2021. The zebrafish developmental toxicity test is being developed as an alternative method in line with the 3R principle from an animal welfare perspective in Japan. However, several issues remain, and attempts are being made to overcome these issues through digital transformation (DX). For example, automated time-lapse imaging for high-throughput, improved accuracy of digital image analysis, quality assurance of fertilized zebrafish embryo used in embryotoxicity tests, and enhancement of accuracy by a breeding quality control cloud system are becoming evident. However, establishing clinical extrapolation, which is the most important aspect of zebrafish embryotoxicity testing, is still a remaining difficulty. To overcome this problem, we have been analyzing the similarity in the correlation between the phenome (morphological abnormalities) and transcriptome in human and zebrafish developmental toxicity, based on the high disease genome homology between zebrafish and humans. It is expected to be a systems toxicology model for analyzing the correlation between structural and physicochemical properties of chemicals and developmental toxicity, and may provide a basis for elucidating the molecular mechanism of developmental toxicity and assessing the risk to humans*. Moreover, digital data of phenotype and transcriptome in developmental toxicity studies and LSKB, a big database, were analyzed by BERT for the first time, a natural language processing model AI developed by Google in 2018. We report on the usefulness of BERT in the zebrafish developmental toxicology. *Congenital Anomalies 2016:56,18-27
