Machine learning for gene regulation in single cells
I build machine-learning models of how transcription factors control cell identity, and I train and test them on single-cell multi-omics data from human immune cells.
Open to research roles in perturbation modeling and foundation models for single-cell data. I expect to finish my PhD in 2027.
What I work on
Perturbation modeling
My current project models transcription-factor perturbations. Details will follow with the preprint.
ARTEMIS
A GATv2 model in PyTorch Geometric that predicts cell type from methylation signatures over a graph of known TF–TF interactions. Its attention weights name the regulators behind each prediction. I am the sole developer.
methylTFR
Transcription-factor activity from DNA methylation. First-author bioRxiv preprint (2026). The annotation packages are accepted to Bioconductor.
Immune-cell epigenomes
scATAC-seq, snmC-seq2 and RNA-seq of human immune cells after viral and chemical exposure. First-author preprint (2025) and co-author of the related Nature Genetics study (2026).
LLMRS
Co-author of a zero-shot recommender system built on a pre-trained large language model.
Tools
Open-source software
I write and maintain three genomics packages: deconvR on Bioconductor, methylTFR and ChrAccR, including its allele-specific accessibility module. Install instructions are on the Software page.
I’m a member of AI Grid, a German network that connects early-career AI researchers with mentors from academia and industry.
