AI/ML Now: Toxicity Prediction for Unknown LC–HRMS Features Using Machine Learning

At HTC-19 in Leuven, Belgium, Viktoriia Turkina (University of Amsterdam) discussed her group’s strategy for prioritizing liquid chromatography–high-resolution mass spectrometry (LC-HRMS) features by predicted toxicity without first identifying the compounds.¹ She explained that confident identification is the main bottleneck in non-target analysis, demanding significant time, resources and expertise. Prioritization narrows the focus to the most relevant features, and because toxicity is a key property in environmental studies, the aim was to steer analysis towards the compounds most likely to harm environmental or human health. With resources limited, this helps labs direct their effort to the most significant findings and get the fastest response for the least resource use.