- Development of an AI-based model to predict the developmental and reproductive toxicity of chemicals, serving as an alternative to animal testing

- Expected applications for next generation risk assessment of environmental chemicals and safe by design of chemicals 

A research team led by Professor Jinhee Choi of the School of Environmental Engineering at the University of Seoul has developed an artificial intelligence–based (AI-based) model for predicting developmental and reproductive toxicity, demonstrating its potential for screening of chemicals found in consumer products. 

▶ Workflow for the development and application of machine learning models to screen the developmental and reproductive toxicity potential of chemicals in consumer products

The study was conducted with Dong-Hyeon Kim, a doctoral student in the School of Environmental Engineering, and Siyeol Ahn, a master’s student, who served as first and second authors, respectively. The results were published online on June 18, 2025, in Environment International—a leading international journal in the field of environmental science (Impact Factor = 9.7 according to 2024 JCR, placing it in the top 7% of the field)—under the title “Identification of developmental and reproductive toxicity of biocides in consumer products using ToxCast bioassays data and machine learning models.”

Humans can be exposed to consumer product chemicals through various routes in daily life, which may cause a range of environmental diseases. However, the current regulatory framework for developmental and reproductive toxicity testing relies heavily on animal studies, which are labor-intensive, require highly specialized skills, and can only test a limited number of chemicals. This has resulted in a consistent call for alternative methods capable of rapidly and efficiently assessing developmental and reproductive toxicity of chemicals.

To address this need, the research team developed a machine learning–based AI model that learns molecular properties and structural information from in vitro cellular and molecular-level data collected from the ToxCast database developed by the U.S. Environmental Protection Agency (EPA). The model successfully predicted the developmental and reproductive toxicity of chemicals contained in consumer products and is expected to serve as an alternative test method to minimize animal testing. It also has the potential to contribute to the increasingly stringent management of chemical safety.

Professor Choi, the corresponding author, stated, “Amid growing public concern over the safety of consumer product chemicals, insufficient hazard assessment remains a problem. To address this, it is essential to develop new approach methodologies (NAMs) by leveraging toxicity big data and AI. Although this research is in its early stages, it represents an important technology for protecting public safety and could be applied in a variety of fields in the future.”

This work was supported by Korea Environmental Industry & Technology Institute (KEITI) through 'Core Technology Development Project for Environmental Diseases Prevention and Management' (2021003310005) and ‘Technology Development Project for Safety Management of Household Chemical Products’ (RS-2023-00215309), funded by Korea Ministry of Environment.

▶ Dong-Hyeon Kim, Siyeol Ahn