AI Model Predicts Developmental Toxicity with Zebrafish Embryo Analysis
MUNICH, GERMANY – A new artificial intelligence model is demonstrating the ability to identify developmental abnormalities in zebrafish embryos, offering a possibly faster and more efficient method for predicting chemical toxicity. Researchers at the CISPA Helmholtz Center for Information Security, led by Sarath Sivaprasad, have developed the system, detailed in a paper presented at the 28th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) in 2025.
The model mimics the analytical capabilities of human experts,analyzing zebrafish progress over time to recognize anomalies and enable early predictions of potential toxicity. Currently focused on a single chemical, the team aims to expand the model’s capabilities to encompass a extensive library of chemicals, creating a valuable resource for both biomedical and AI research. the complete data record generated by the project is freely available on GitHub,encouraging community collaboration and the development of more advanced,efficient,and ethical toxicity screening methods.
Zebrafish embryos are frequently used in developmental biology research due to their openness, rapid development, and genetic similarity to humans. Traditionally, assessing the impact of chemicals on these embryos requires painstaking manual observation. This new AI-driven approach automates the detection process, potentially accelerating the identification of harmful substances.
“The data record is a valuable resource for both the machine learning community to evaluate its methods, and also for biomedical research to better understand the effects of different active ingredients,” explained Sivaprasad.The research team,which includes Hui-Po Wang,Anna-Lisa Jäckel,Jonas Baumann,Carole Baumann,Jennifer Herrmann,and Mario Fritz,hopes the open-source data will foster innovation in toxicity screening and contribute to safer chemical development.
