Our Research

Analysis of large and complex datasets requires scalable and reproducible computational methods. We have a track record of developing computational tools, including the simulation engine AMICI [Fröhlich et. al 2021] and the calibration toolbox pyPESTO [Schälte et. al 2023], and strive to continue to push the envelope through methodological innovation.

Mathematical models are powerful means to encode prior knowledge. While valuable in data-sparse settings, this limits unbiased discovery. We apply data-driven methods to enable the unbiased inference of explainable dynamic models from large perturbational screens.

Signalling exhibits considerable cell-to-cell variability resulting in heterogeneous cellular decision-making outcomes. We build models of signal transduction that predict these outcomes based on quantifications of the molecular make-up of cells [Fröhlich et. al 2018].

Signalling is intricately regulated by numerous molecular mechanisms involving protein-protein interactions and post-translational modifications. We build detailed mechanistic models that account for protein structure [Gerosa et. al 2020, Fröhlich et. al 2023] to learn the emergent regulatory principles for signalling.