The safety of water, food, and new materials used in daily life is essential for human life and the ecosystem. While nontarget liquid chromatography high resolution mass spectrometry (LC/HRMS) is increasingly used to detect chemicals in such samples, evaluating the hazard possessed by the chemicals in these complex mixtures, especially endocrine disruptive potency, still requires animal experiments. 

Evaluating the hazard of chemicals present in complex mixtures does not have to come with the expense of animal welfare, as todays’ computational tools such as artificial intelligence and in silico deep learning algorithms represent a promising alternative. 

However, a major limitation that hinders the full-scale application of these methods for predicting toxic endpoints of complex mixtures has been that single chemical constituents in the mixture first need to be unequivocally identified. 

Here Anneli and team propose that recent developments in molecular networking, semi-supervised learning, and deep learning allow predicting the hazard of chemicals directly from the empirical spectral information acquired rapidly in nontarget LC/HRMS chemical analysis.