Squeaky clean learning

Ever wondered what goes into making your shampoo? It’s not just about making your hair squeaky clean – it’s a bubbly mix of science and suds!

Customers' needs have evolved over the years and there is now a greater demand than ever before for shampoos that are eco-friendly, and swap out fossil carbon-based ingredients for natural ones. But getting the perfect formula is like trying to untangle the worst hair knot ever – it’s super tricky and takes a lot of time.

So, to wash away these troubles, scientists are turning to machine learning (ML) and generation of experimental data by robots. They hope that these high-tech helpers can predict the best ingredient mixes. In a previous study, a small ML model to help design shampoos was created but it needed more data – it was like trying to wash long hair with a tiny dollop of shampoo!

In this new study, published in international scientific journal Scientific Data (Nature), the focus is on creating a large, open dataset for training ML models specifically for shampoo. They picked 12 different surfactants (the cleaning agents), four conditioning polymers (to keep hair smooth and silky), and two thickeners (to get that perfect consistency).

Mixing these ingredients is like a shampoo cocktail party, with 528 unique combinations – way more than before. And to make this possible an international research team has designed new formulation robots, created intelligent algorithms to design experiments and run a highly intensive campaign to record the data. Now anyone can have a go at creating a predictive model of a shampoo formulation.

So, the next time you reach for the shampoo, take a moment to think of the work it took by our Department of Chemical Engineering and Biotechnology (CEB) colleagues.

CEB PhD student Aniket Chitre worked on this project conceived by Professor Alexei Lapkin and colleagues at BASF Advanced Chemicals Co (Shanghai). They teamed up with experts in high-throughput experiments, the research group of Professor Kedar Hippalgaonkar of Nanyang Technological University (Singapore), and were helped by many researchers in the international team: Simon Rihm (CEB), Dogancan Karan (Cambridge CARES), Robert Querimit (Nanyang Technological University), Benchuan Zhu, Ke Wang and Long Wang of BASF Advanced Chemicals Co.

Read the study: Accelerating Formulation Design via Machine Learning: Generating a High-throughput Shampoo Formulations Dataset | Scientific Data (nature.com)