AI-driven approach reveals hidden hazards of chemical mixtures in rivers

A new approach, developed by researchers at the University of Birmingham, demonstrates how advanced artificial intelligence (AI) methods can help identify potentially harmful chemical substances in rivers by monitoring their effects on tiny water fleas (Daphnia).

The team worked with scientists at the Research Centre for Eco-Environmental Sciences (RCEES), in China, and the Hemholtz Centre for Environmental Research (UFZ), in Germany, to analyse water samples from the Chaobai River system near Beijing. This river system is receiving chemical pollutants from a number of different sources, including agricultural, domestic and industrial.

Professor John Colbourne is the director of the University of Birmingham’s Centre for Environmental Research and Justice and one of the senior authors on the paper. He expressed optimism that, by building upon these early findings, such technology can one day be deployed to routinely monitor water for toxic substances that would otherwise be undetected.

He said: “There is a vast array of chemicals in the environment. Water safety cannot be assessed one substance at a time. Now we have the means to monitor the totality of chemicals in sampled water from the environment to uncover what unknown substances act together to produce toxicity to animals, including humans.”

The results, published in Environmental Science and Technology, reveal that certain mixtures of chemicals can work together to affect important biological processes in aquatic organisms, which are measured by their genes. The combinations of these chemicals create environmental hazards that are potentially greater than when chemicals are present individually.

The research team used water fleas (Daphnia) as test organisms in the study because these tiny crustaceans are highly sensitive to water quality changes and share many genes with other species, making them excellent indicators of potential environmental hazards.

“Our innovative approach leverages Daphnia as the sentinel species to uncover potential toxic substances in the environment,” explains Dr Xiaojing Li, of the University of Birmingham (UoB) and the lead author of this study. “By using AI methods, we can identify which subsets of chemicals might be particularly harmful to aquatic life, even at low concentrations that wouldn’t normally raise concerns.”

Dr Jiarui Zhou, also at the University of Birmingham and co-first author of the paper, who led the development of the AI algorithms, said: “Our approach demonstrates how advanced computational methods can help solve pressing environmental challenges. By analysing vast amounts of biological and chemical data simultaneously, we can better understand and predict environmental risks.”

Professor Luisa Orsini, another senior author of the study, added: “The study’s key innovation lies in our data-driven, unbiased approach to uncovering how environmentally relevant concentrations of chemical mixtures can cause harm. This challenges conventional ecotoxicology and paves the way to regulatory adoption of the sentinel species Daphnia, alongside new approach methodologies.”

Dr Timothy Williams of the University of Birmingham and co-author of the paper also noted that: “Typically, aquatic toxicology studies either use a high concentration of an individual chemical to determine detailed biological responses or only determine apical effects like mortality and altered reproduction after exposure to an environmental sample. However, this study breaks new ground by allowing us to identify key classes of chemicals that affect living organisms within a genuine environmental mixture at relatively low concentration while simultaneously characterising the biomolecular changes elicited.”

The findings could help improve environmental protection by:

• Identifying previously unknown chemical combinations that pose risks to aquatic life
• Enabling more comprehensive environmental monitoring
• Supporting better-informed regulations for chemical discharge into waterways

This research was funded by the Royal Society International Collaboration Award, the European Union’s Horizon 2020 research and innovation programme, and the Natural Environmental Research Council Innovation People programme.