University of Tübingen Study Challenges Standard Bioaccumulation Testing and Introduces AI-Based Solution

Rita Triebskorn and Heinz Köhler working with the AI-based BCFpro program, which theoretically determines the bioconcentration factor (BCF) in relation to chemicals in fish.

(IN BRIEF) Researchers from the University of Tübingen have identified a major weakness in current methods used to assess the bioaccumulation of chemicals in fish, showing that the bioconcentration factor varies with test concentration rather than being a fixed substance-specific value. This finding casts doubt on regulatory data used for many chemicals in the EU. In response, the team developed an AI-based tool, BCFpro, which can predict bioaccumulation with high accuracy and identify worst-case scenarios. Made freely available, the tool promises more reliable risk assessments, improved environmental safety, and a significant reduction in animal testing.

(PRESS RELEASE) TÜBINGEN, 13-Jan-2026 — /EuropaWire/ — A research team led by the University of Tübingen has uncovered a fundamental flaw in the way bioaccumulation of chemicals in fish is currently assessed, calling into question a large body of data used in regulatory decision-making across Europe. The interdisciplinary group has shown that the bioconcentration factor (BCF), a key metric used worldwide to evaluate whether chemicals accumulate in aquatic organisms, is not a fixed value for individual substances as previously assumed. Instead, it varies depending on the concentration applied during testing. To address this issue, the team has developed a new artificial intelligence-based tool that enables far more reliable assessments of bioaccumulation and is being made freely available to the scientific community.

The bioconcentration factor compares the concentration of a chemical in fish with that in the surrounding water and plays a central role in environmental risk assessments, including EU chemical licensing procedures. By analysing thousands of published studies, the research team demonstrated that higher test concentrations in water typically lead to lower measured BCF values, while lower test concentrations tend to produce higher BCFs. This systematic bias had not previously been identified or accounted for in chemical hazard classification regulations, despite its significant implications for environmental and human health.

The findings suggest that bioaccumulation data for more than half of the chemicals currently considered capable of accumulating in fish may be unreliable. Because bioaccumulated substances can move up the food chain and ultimately affect humans, inaccurate classification poses long-term risks that may only become apparent years later. The researchers were able to demonstrate the concentration-dependent effect both mathematically and physiologically, highlighting a critical gap between regulatory practice and biological reality.

Building on these insights, the team applied deep learning techniques to develop an AI-based program capable of predicting bioconcentration factors with around 90 percent certainty. The tool, named BCFpro, uses artificial neural networks to process complex datasets and identify patterns that traditional methods overlook. In addition to predicting experimental outcomes, BCFpro can model worst-case scenarios, identifying conditions under which chemicals are most likely to bioaccumulate at dangerous levels.

When benchmarked against existing regulatory classifications, the new tool confirmed established results for substances already recognised as bioaccumulating in around 90 percent of cases. However, it revealed a far more concerning picture for chemicals currently classified as non-bioaccumulative: more than 60 percent of these substances should, under environmentally relevant conditions, have been flagged as bioaccumulating. According to the researchers, this discrepancy arises because previous test conditions were often chosen in ways that underestimated worst-case accumulation.

By making BCFpro freely accessible, the University of Tübingen-led team aims to improve the accuracy and consistency of chemical risk assessments worldwide. Beyond strengthening environmental protection, the AI-based approach also offers significant potential to reduce animal testing by reliably predicting bioaccumulation through computational methods. The researchers emphasise that their work not only challenges long-standing assumptions in ecotoxicology but also contributes to safer environmental practices and improved animal welfare.

More Information:

AI tool BCFpro for standardized and reliable categorization of chemicals: https://www.parcopedia.eu/

Publication:

Heinz-R. Köhler, Reza Aalizadeh, Gabriele Treu, Thomas Gräff, Katharina Peschke, Ines Prutz, Nikolaos S. Thomaidis, Rita Triebskorn, Peter C. von der Ohe: By integrating previously overlooked drivers AI boosts bioaccumulation assessment in fish. Journal of Hazardous Materials, https://doi.org/10.1016/j.jhazmat.2025.140648

Contact:

Prof. Dr. Heinz R. Köhler
University of Tübingen
Institute of Evolution and Ecology (EvE) – Animal Physiological Ecology
Phone +49 7071 29-78890
heinz-r.koehler@uni-tuebingen.de

Contact for press:

Eberhard Karls Universität Tübingen
Public Relations Department
Christfried Dornis
Director

Janna Eberhardt
Research Reporter
Phone +49 7071 29-77853
janna.eberhardt@uni-tuebingen.de

SOURCE: University of Tübingen