Auditing single-cell perturbation benchmarks for leakage and confounding
A short note on why PerturbGuard treats single-cell perturbation benchmarks like data-quality objects before model results are interpreted.
Blog
Short write-ups from selected computational biology projects: what I tried, what the data or workflow showed, what I would trust, and what still needs work.
A short note on why PerturbGuard treats single-cell perturbation benchmarks like data-quality objects before model results are interpreted.
A short note on building a human adenovirus workflow that goes from raw reads to assemblies, consensus genomes, variability plots, and phylogeny.
A short note on building StrataDock as a docking workstation where receptor preparation, pocket choice, docking, validation, and reporting stay connected.
A short note on testing whether zero-shot ESM scores recover useful signal from APP deep-mutational scanning data.
A short note on using methylation clocks, immune deconvolution, RNA expression, and clinical metadata to study PAM50 breast cancer subtype signal.
A short note on designing a bacterial whole-genome sequencing workflow that connects reads, assemblies, annotations, screening, and phylogeny.