Back to Case Studies
Human scRNA-seq GSE85241 Scanpy Leiden + UMAP

Single-Cell Clustering and Marker Analysis of Human Pancreatic Islets

A first-party account of the Scanpy session: count-integrity checks, quality control, donor and plate assessment, Leiden clustering, Wilcoxon marker testing, working cell-type annotations, and the limitations that constrain interpretation.

Inspect the findings Download session PDF
Input cells3,072
After QC2,681
Leiden clusters18
Donors4
01 / Scientific question

Which transcriptionally distinct populations are present in the pancreatic-islet dataset?

The session analyzed GSE85241 as a cell-atlas-style dataset. Its goal was cluster discovery and marker identification—not a treatment-versus-control comparison. The four donor and plate identifiers encoded in the cell barcodes were reconstructed and assessed before clustering.

Source dataset
NCBI GEO GSE85241
Input matrix
19,140 genes × 3,072 CEL-seq2 cells
Sample structure
4 donors × 8 plates × 96 wells
Analysis objective
QC, clustering, working annotations, and per-cluster markers
02 / Managed execution

Workflow and decision gates

The report does more than enumerate tools: it records why normalization and batch-integration decisions were made.

Count-integrity gateNon-integer values were identified as CEL-seq2 UMI-collision-corrected transcript counts and treated as count-like data.
Quality controlCells with fewer than 200 detected genes and genes present in fewer than 3 cells were filtered; 2,681 cells and 17,792 genes remained.
Normalize and reduceScanpy normalization, log1p, 2,000 highly variable genes, scaling, PCA, a 20-neighbor graph, and UMAP.
Assess batch structureDonor silhouette score on the top PCs was −0.071, so Harmony or BBKNN integration was not applied.
Cluster and annotateLeiden at resolution 1.0 produced 18 clusters; Wilcoxon marker tests and a pancreas marker dictionary produced working annotations for review.
03 / Results

Major endocrine and exocrine populations were resolved—with one clear technical artifact

Marker profiles supported working annotations for alpha, beta, delta, PP/gamma, acinar, ductal, stellate, and endothelial populations. The run retained 36,833 significant marker gene–cluster associations after Bonferroni correction.

Expected populations

Large clusters carried canonical markers including INS, SST, PPY, PRSS1, CFTR, SPARC, and PLVAP.

Artifact cluster

Cluster 6 contained 247 low-complexity cells dominated by ERCC spike-in transcripts and was not treated as a real delta-cell population.

Rare-cell hypothesis

Cluster 13 contained 10 KIT/CPA3/TPSB2-positive cells consistent with a candidate mast-cell population, pending validation.

Batch decision

Donor identity did not dominate the leading PCs, so clustering used the uncorrected PCA and neighbor graph.

UMAP of 2,681 quality-controlled human pancreatic islet cells colored by 18 Leiden clusters
Report figure: UMAP of the 2,681 QC-passed cells colored by Leiden cluster. Cluster separation alone does not validate the working cell-type labels.
04 / Interpretation boundary

Limitations are part of the result

  • The input contains non-integer, UMI-collision-corrected transcript counts rather than raw integer UMIs.
  • No mitochondrial genes were present in the deposited annotation, so mitochondrial-fraction QC could not identify damaged cells.
  • Cluster 6 is ERCC-dominated and low-complexity; clusters 10, 13, and 17 are too small or ambiguous for confident biological interpretation.
  • The cell-type labels are marker-supported hypotheses, not identities validated by protein measurements, spatial data, or an external pancreas reference.
  • GSE85241 has no treatment/control design, so this analysis does not establish condition-level differential expression.
Confidence boundary

Moderate-to-high for the major islet and exocrine working annotations; low for very small clusters and fine subtype distinctions.

05 / Provenance

Reproducibility artifacts recorded in the session

The report inventories the AnnData checkpoints, result tables, QC flags, figures, and captions used to support the interpretation.

adata_raw.h5adLoaded matrix with donor and plate metadata
adata_qc.h5ad2,681 cells × 17,792 genes after QC
adata_clustered.h5adLeiden clusters and graph embeddings
cluster_markers.csvSignificant marker gene–cluster associations
cluster_annotations.csvCluster sizes, scores, and working labels
analysis_flags.jsonMachine-readable input and interpretation caveats
Original session report20-page PDF containing the methods, QC accounting, cluster table, limitations, output inventory, and figures.
Download PDF

Related Case Studies

Single-Cell
Lung Cancer TME Literature Scan
RNA-seq
Rice Salt-Alkali Stress DE Analysis
RNA-seq
Dexamethasone Airway Smooth Muscle DE Analysis

Run a similar single-cell analysis

Bring your matrix and biological question to a managed workspace with explicit QC and interpretation gates.

Start an analysis