perturbation_de_vs_control

Interpret a VCF's variants (germline/somatic, WGS/WES) — gene + functional consequence (Ensembl VEP), ClinVar pathogenicity, gnomAD frequency, deleteriousness…

Analyze a pooled CRISPR / Perturb-seq screen — rank which perturbations change the transcriptome, then per-perturbation DE vs the non-targeting control

scGPT foundation-model per-cell annotation for a .h5ad — transfers reference cell-type labels + calibrated confidence to every cell, AND independently…

Assign a cell-type label to each cluster from a curated marker panel, using signature scoring for a first pass and RAW marker expression to make the final call.

Reference template — normalize a VCF (split multiallelic, atomize MNPs, left-align indels) with bcftools norm before annotation or ClinVar/gnomAD matching.

Reference template — score every cell for one or more gene signatures with scanpy `sc.tl.score_genes` (signature mean minus a matched random control set)…

Compare an experimental condition vs control (e.g.

Reference template — prioritize disease-associated genes by combining condition differential expression with a precalculated cell-type atlas.

Reference template — assign a research-triage priority tier (pathogenic → VUS → benign) per variant from ClinVar + gnomAD AF (disease-model thresholds) +…

Single-cell cell-type annotation + report (v2 — stability-selected resolution, raw-expression label confirmation, preranked GSEA)

Reference template — rank perturbations by E-distance to control (which guides actually did something).

Reference template — condition-vs-control differential expression, STRATIFIED BY CELL TYPE.