Bayesian modelling for optimization of untargeted

This project aims to optimize experimental design in untargeted metabolomics by combining design of experiments (DoE) with Bayesian multilevel modeling. We analyze large-scale LC–MS datasets to quantify how dilution factor and injection volume affect both signal intensity and measurement variability across hundreds of metabolic features. The goal is to identify optimal acquisition settings that maximize sensitivity and reproducibility.

Untargeted metabolomics generates high-dimensional datasets with substantial variability arising from both biological and technical sources. Experimental design choices, such as dilution factor and injection volume, critically influence signal intensity and measurement precision, yet their joint effects are not well characterized.

The main objective is to derive statistically robust, data-driven recommendations for experimental settings that improve quantitative performance in metabolomics workflows.