Perturbation-aware representation learning
Representation-learning approaches for heterogeneous single-cell screens, designed to preserve perturbation-relevant gene-level signal and resolve weak effects.
Methods & software
Reusable computational methods
Each method grew from a specific experimental or analytical constraint, but was designed as a general approach rather than a one-off analysis.
Active research
Record-seq analysis framework
An end-to-end computational workflow for transcriptional-recording data, from raw sequencing reads through quality control, statistical testing and downstream interpretation.
Record-seq gut sensing
Analysis workflows for transcriptional recording sentinel cells used in non-invasive assessment of gut function, spanning the foundational Record-seq protocol and its biological application.
WiNN
A white-noise-normalisation method for correcting run-order drift and batch effects in metabolomics while retaining biological signal.
Differential de Bruijn assembly
A scalable approach to differential sequence assembly in metagenomic data using annotated de Bruijn graphs.