Machine Learning for Computational Water Treatment

Endocrine-disrupting chemicals (EDCs) affect the hormone systems of animals, mimicking the effects of naturally occurring hormones (such as oestrogen or testosterone) in animals and blocking their action. The effects of these chemicals are wide-ranging and include reproductive failure and developmental problems. Unfortunately, a huge variety of compounds have the potential to disrupt the endocrine system, including pharmaceuticals (e.g. antibiotics), personal care products (e.g. deodorants) and raw materials for manufacturing (e.g. bisphenols). While the full effect of these compounds on human health is not yet known, their removal from drinking water is an emerging problem in water treatment, and one that only becomes more important as more EDCs are discovered. The best way to remove a given EDC from drinking water is not always obvious, and the standard practice is to screen different possible methods to find the optimum. This can be very costly in terms of both money and time, and the method that is best for one source of drinking water may not always be best in another source whose composition is different.

This project harnesses the power of computational chemistry and machine-learning (ML) to speed up the search for materials for EDC removal, beginning with atomistic simulations to study water decontamination in silico, in tandem with the results of laboratory experiments. The culmination of this work will be the development of an efficient and robust ML framework that can predict the ability of a material to remove an endocrine disruptor from drinking water, saving a significant amount of experimental time by suggesting candidate materials to focus on, and allowing the water management industry to act quickly to deal with newly discovered EDCs.