A synthetic MALDI-TOF MS atlas for microbial classification across hospitals
Preprint on bioRxiv, 2026
Recommended citation: Guerrero-López, A., Meijer, E., Gütlin, Y., Koch, K., Rabbani, N., Hitz, E., Serrano-Conde, E., Howden, B. P., Søgaard, K. K., Syed, M. A., Oberle, M., Moran-Gilad, J., Hernández-García, M., Sherry, N. L., & Egli, A. (2026). A synthetic MALDI-TOF MS atlas for microbial classification across hospitals. bioRxiv [Preprint]. https://doi.org/10.64898/2026.10.19.759651 https://www.biorxiv.org/content/10.64898/2026.10.19.759651v1
bioRxiv preprint, posted on 19 October 2026 (version 1).
My role: first and corresponding author. This work forms part of my Marie Skłodowska-Curie Postdoctoral Fellowship at the University of Zurich, supported by MSCA grant 101278403 (OUTBRAID).
We created an open atlas of 4.3 million synthetic MALDI-TOF mass spectra covering 43 clinically relevant bacterial species, using a generative model trained on 179,086 measured spectra. Models trained on the synthetic atlas identified bacteria in independent hospital datasets, and adding synthetic spectra matched or improved performance compared with training on measured data alone.
The work explores how synthetic spectra can help develop microbial classification models that transfer across hospitals, addressing differences between the datasets collected at different institutions.
Read the preprint · DOI · Code