Involved versus uninvolved Crohn ileum - donor-aware expression assessment

Research question

Which expression contrasts remain credible when paired donors, rather than individual cells, define replication?

Use public GSE134809 single-cell data to compare involved and uninvolved Crohn ileal tissue across 11 paired donors. The task asks for donor-supported inflammatory expression differences, a clean computation from frozen inputs, inspectable analysis artifacts, and careful treatment of tissue definitions, donor replication, chemistry, annotations and statistical limitations. A within-cohort expression contrast is not independent biological validation, a healthy-control comparison, causal evidence or a treatment-response prediction.

Data and comparison design

Cells from the same donor share biological and technical context. Read the paired-donor contrast separately from the number of cells, and distinguish the composite association rule from independently corrected panel-gene and program tests. The V2-only sensitivity changes the chemistry subset.

Input or source	Role in this case	Source context
GSE134809	Public Crohn ileal single-cell source data	Credit GEO and the original study; use the frozen input hashes for exact reconstruction
Martin et al., Cell 2019	Study design and biological context	The publication's cohort and annotations are not interchangeable with this reanalysis

Recorded findings

Testing layer	Recorded result	Replication unit or scope
Input	69,762 cells from 11 paired donors	Donors, not cells, define independent biological replication.
Composite association rule	15 inflammatory gene–lineage associations	Separate from the individual-gene testing family.
Individual panel genes	None pass separately corrected exact-test FDR	A stricter, separately corrected evidence layer.
Program tests	6/18 pass primary FDR; none in V2-only sensitivity	Chemistry restriction changes the support.

The selected Session was exported on 2026-09-29 from a new clean project. It records a gpt-6-astra run with 36 Notebook executions, including two failed executions retained in the history, and seven Artifact lineages / eight Artifact versions. The saved report describes 69,762 assigned immune/stromal cells from 22 ileal libraries and 11 paired donors. Six lineages have 8–11 eligible donor pairs. It reports 15 inflammatory gene–lineage associations meeting its composite negative-binomial/donor-consistency rule, strongest in myeloid cells; none of the individual inflammatory-panel genes crosses the separately corrected exact-test FDR threshold. Six of 18 primary program tests narrowly cross FDR <0.05, while none does in the V2-only sensitivity. These are the producing Session's reported results, not calculations independently recomputed for this repository submission.

What the comparison shows

Many cells do not replace independent donors. Different testing families and chemistry restrictions can produce different discovery counts without being contradictory.

Assumptions and limits

Small donor counts, chemistry and annotation choices constrain this within-cohort contrast. Historical review issues were addressed in review_dispositions.json; textual review does not establish a full computational audit or fresh-environment Notebook rerun.

Sources and research materials

The unchanged research package comes from upstream Case 12. It includes the archived session, analysis materials, results and figures. The poster uses an AI-generated illustrative photograph. This introduction summarizes archived findings; the analysis was not rerun.