Lilian Gasser

This study trains and evaluates machine learning models to predict fish acute mortality from chemical properties, finding that tree-based models perform best but struggle with individual chemical predictions.

The paper introduces ADORE, a benchmark dataset for machine learning in ecotoxicology, to facilitate model comparison and challenge researchers to improve predictive performance.

The paper evaluates machine learning methods for predicting hydrophobicity as a proxy for retention time in tandem mass spectrometry, finding that ensemble methods and multitask learning improve generalization.