AI Revolutionizes Environmental Risk Assessment from Chemicals
The bioconcentration factor (BCF) is a critical measure of chemical substances in fish compared to the surrounding water, indicating bioaccumulation in the environment. Until recently, it was believed to be a constant for each chemical. However, groundbreaking research led by Professor Heinz Köhler from the University of Tübingen's Institute of Evolution and Ecology challenges this notion. The study reveals that BCF varies depending on the test concentration, casting doubt on the accuracy of bioaccumulation data used in the EU's licensing procedure for over half of the chemicals that potentially accumulate in fish.
This discovery has significant implications, as chemical concentrations in the food chain, including humans, can lead to harmful build-up in the body over time. Professor Köhler emphasizes the importance of accurate BCF assessments, stating, 'Concentrations can build up massively in the human body, and harmful effects may only become apparent after a long time.'
To address this issue, the research team has developed an innovative AI tool, BCFpro, which enables researchers to assess the bioaccumulating properties of substances with remarkable precision. This tool is freely available, and the team's findings are published in the Journal of Hazardous Materials.
The team's research involved evaluating thousands of studies on chemical tests, revealing that the BCF factor is not specific to each chemical. The concentration of the test substance in the surrounding water significantly influences the BCF, with higher concentrations generally resulting in lower BCF values. This finding was mathematically proven and physiologically explained by the team, marking a significant advancement in chemical hazard classification regulations.
To enhance efficiency, the team employed deep learning, an AI machine learning method, to develop a program that can predict experimental BCF data with 90% certainty. This method utilizes artificial networks, mimicking brain neurons, to process complex datasets and identify patterns. The tool can also identify critical values for chemicals under worst-case scenarios, ensuring more accurate risk assessments.
When applying the new tool to substances categorized as bioaccumulating in the EU, the team achieved 90% accuracy. However, a concerning revelation emerged when reviewing substances categorized as non-bioaccumulating: over 60% were incorrectly identified. This highlights the importance of conducting chemical tests under environmentally relevant conditions to obtain realistic risk assessment values.
The research team's commitment to standardized and reliable chemical categorization led to the development of BCFpro, which is now freely available. This AI tool has the potential to significantly reduce animal testing by reliably predicting the bioaccumulation of new chemical developments. Professor Karla Pollmann, President of the University of Tübingen, praises the study's practical focus, emphasizing its contribution to improved ecotoxicological methods and environmental safety.
