v0.2.0 — Pathway-level analysis, hardened and packaged

ncountr 0.2.0

This release brings pathway-level analysis to ncountr and hardens the package with a comprehensive, multi-version test suite. It also corrects the package version: 0.1.0 was the last release on PyPI, but the code already self-identified as 0.2.0 — this release reconciles that.

nCounter studies usually have small sample sizes (n=3–12) where gene-level FDR correction returns nothing significant. The headline of 0.2.0 is a set of pathway-level methods built for exactly that regime. Those analysis features landed during development (commit 9c84e3d); the work since has been test hardening, bug fixes, and packaging correctness.

✨ New analysis features

• Permutation-based GSEA (Subramanian et al.) with exact enumeration for small n (≤10 total samples), plus a pre-ranked variant — gsea, gsea_preranked.
• Competitive gene-set test (CAMERA-like): Wilcoxon rank-sum on per-gene t-statistics, in-set vs out-of-set — competitive_test.
• Self-contained permutation test on pathway-score differences — self_contained_test.
• 12 built-in MSigDB Hallmark gene sets for immune/host-response panels, plus a GMT parser for custom sets — load_gmt / save_gmt.
• Effect-size reporting: Cohen's d with bootstrap CIs and rank-biserial correlation — effect_sizes, de(effect_size=True).
• ssGSEA single-sample pathway scoring.
• New figures: GSEA enrichment plot, GSEA dot plot, and effect-size volcano.

✅ Reliability & testing

• Test suite expanded from 150 → 291 tests, now covering every module (GSEA, pathway, effect sizes, gene sets, datasets, export/AnnData, GEO download, CLI, plotting, cross-platform). Coverage ~93%.
• CI runs the full matrix on Python 3.9, 3.10, 3.11, 3.12 and installs the crossplatform extra, so the AnnData/scanpy code paths are actually exercised.

🐛 Bug fixes

• GSEA empty result: competitive_test, self_contained_test, and gsea_preranked crashed with KeyError: 'pvalue' when no gene set passed the overlap filter — common on small panels. They now return a correctly-typed empty result.
• matplotlib ≥3.11: the pathway box plot crashed because boxplot(labels=…) was removed. Now uses set_xticklabels, compatible with matplotlib ≥3.5.
• Python 3.9 CLI: ncountr de failed with "No such command 'de'" under Click 8.1 (newest on 3.9), which doesn't strip the _cmd suffix. The command name is now pinned explicitly.

📦 Packaging

• Version is single-sourced from pyproject.toml (read at runtime via importlib.metadata), resolving the 0.1.0/0.2.0 mismatch so built artifacts report the correct version. A test guards against future drift.

Full changelog: v0.1.0...v0.2.0

v0.1.0 — Initial release

ncountr v0.1.0

First public release of ncountr — the first Python package for end-to-end Nanostring nCounter gene expression analysis.

Features

• RCC parsing — read .RCC files from one or more directories into a structured experiment object
• Quality control — FOV ratio, positive control linearity, housekeeping stability, negative background
• Normalization — positive control, housekeeping, and background subtraction methods
• Differential expression — Mann-Whitney U or t-test with FDR correction
• Gene set scoring — built-in IFN/JAK-STAT pathway (48 genes) + custom gene sets
• Publication-ready plots — QC summary, volcano, heatmap, pathway score box plots
• Cross-platform validation — correlation, DE concordance, and composition analysis vs RNA-seq
• AnnData export — to_anndata() for scverse ecosystem integration
• GEO downloader — ncountr fetch-geo GSE275334 downloads RCC files directly
• Config-driven pipeline — single YAML file runs the entire analysis

Validated on 5 published datasets (1,458 samples)

Dataset	Panel	Samples	Key result
GSE275334	Immune Exhaustion	47	Long COVID / ME/CFS 3-group design
GSE140901	PanCancer Immune	24	ICI responder vs non-responder
GSE117751	Human Immunology	42	2 FDR-significant genes (AIR vs Control)
GSE268012	Human Metabolism	24	232 DE genes for IFNβ vs Control
GSE74821	PAM50 Custom	1,321	Parsed in 3s, 99.5% QC pass rate

Install

pip install ncountr