Research questions and methods

Building upon AAV-Perturb-seq, I lead the experimental design for genome-scale perturbation studies in live mice, including target selection and guide design.

I then lead the analysis end to end, from quality control and statistical testing to model evaluation. Where existing tools are insufficient, I develop and apply new statistical methods and machine-learning frameworks to distinguish weak perturbation effects and interpret the structure of complex single-cell data.

As the computational lead for this programme, I am contributing to several related in vivo perturbation manuscripts now in development as a co-first author or co-author.

We have used CRISPR to engineer bacteria into recorders of their own molecular activity. I lead computational method development for this novel data modality and was the computational lead for applying these engineered bacteria as non-invasive reporters in the mammalian gut. This work includes statistical models and analysis workflows, as well as an end-to-end framework for processing transcriptional-recording data.

The work connects synthetic biology with quantitative inference: engineered cells provide a new measurement, and the computational method determines what can be learned from it.

I develop methods for data types whose technical structure can obscure the biological signal of interest, including metabolomics and metagenomics.

Current work includes selective correction of drift and batch effects in metabolomics and differential sequence assembly with annotated de Bruijn graphs.

Within the Swiss Tumor Profiler consortium, I led the generation and analysis of metagenomic and metabolomic data for a metastatic melanoma cohort and contributed to multimodal clinical studies in melanoma and ovarian cancer.

This work connects single-cell, spatial, genomic, proteomic and functional measurements to questions of tumour heterogeneity, drug response and clinical decision support.