MS2Tox: Deep Learning for Automated Prediction of the Endocrine Disruptive Potency of Chemicals in Complex Mixtures

Here we develop MS2Tox, an artificial intelligence for fully automated high throughput read across for the endocrine disruption profile of known and unknown chemicals in complex mixtures.

We anticipate that such workflows can be validated to replace animal tests, such as those that are still suggested for Whole Effluent Assessment in the Baltic Sea Region, which currently utilizes fish and daphniates.

MS2Tox represents an alternative Integrated Approach to Testing and Assessment (IATA) that leverages data-rich fragmentation mass spectra obtained in nontargeted liquid chromatography high resolution mass spectrometry (LC/HRMS) analysis of complex mixtures to link unidentified chemicals to chemicals with known endocrine disruptive potency.

In combination with deep learning, MS2Tox will enable automated and simultaneous prediction of the toxic activity of all detected chemicals towards multiple endpoints and flag chemicals based on the endocrine disruptive potency for mitigation without the need for further animal testing.The project will be performed over three years as a collaboration between a team of expert toxicologists, chemists, and data scientists at Stockholm University and RISE.

The project will build on previously acquired and publicly available bioassay data without requiring additional animal testing.

Finally, Tox21Network will be wrapped into an open-source software package and made universally available to the scientific community and regulators.