5 Plan your analysis
After trying the demo, choose one group of GEM wells to analyze together. This group is an aggregation. It should reflect the biological comparison you intend to make; donor identity, sample preparation, and batch information need to remain distinguishable within it.
5.1 What you need to prepare
- Cell Ranger ARC outputs for each GEM well and the
reference.jsonused to produce them. All GEM wells in an aggregation must use the same reference. - A GEM-well table linking stable identifiers to input paths and library annotations.
- A donor table with one row per donor, including the phenotypes or covariates you will use.
- Marker genes appropriate for the expected cell types, plus initial processing settings that you will review against your data.
Start with the main processing workflow. Donor demultiplexing, CellBender, Harmony batch correction, and downstream modules require additional inputs or choices; enable them when your study needs them.
5.2 How the data move through the workflow
| Stage | What it does |
|---|---|
| GEM well processing | Reads Cell Ranger matrices and fragments, calculates QC metrics, and prefixes barcodes with the GEM well identifier. |
| GEX (gene expression) | Combines selected GEM wells, reduces dimensions, clusters nuclei, and annotates cell types using marker genes. |
| ATAC (chromatin accessibility) | Combines fragments, defines peaks, builds the peak-count matrix, and summarizes accessibility and motif signals. |
| WNN (weighted nearest neighbors) | Combines RNA and ATAC representations to produce integrated clusters, metadata, and a Seurat/Signac export. |
The pipeline also produces comparisons of related GEM wells before aggregation. The main gallery shows representative outputs; the implementation graph provides the detailed computational dependencies when you need them.
Continue to Configuration and inputs to connect your files, then follow the numbered QC reviews.