Cell QC — Barcodes to Cells

Header: 🧪 Cell QC

Available on single-cell RNA-seq projects.

The Cell QC tab answers a single question: how much of your sequencing survived, per sample? It traces every sample from raw barcodes through cell calling, quality filtering, and doublet removal, so you can see where cells were lost before any biology is interpreted.

Example output from the sample project.

Barcode Summary

A summary line at the top reports, for every sample, how many barcodes were observed and how many cells were called from them — for example, "Normal_1: 121,743 barcodes → 9,792 cells called". A large gap between the two numbers is normal: most barcodes are empty droplets.

QC Funnel per Sample

The main chart shows each sample at three stages, so the drop-off between them is visible at a glance:

• cells_called — barcodes judged to contain a cell
• cells_qc_pass — cells surviving the quality filters
• cells_final — cells remaining after doublet removal

The legend on the right identifies samples by color. A sample that falls much further than its peers between two stages is the one to investigate.

Doublets Removed

Below the funnel, a line reports how many doublets were removed from each sample. Doublets are droplets that captured two cells and would otherwise appear as a spurious intermediate cell type.

Per-Sample Medians

An expandable table gives the per-sample quality medians:

Column	Meaning
droplet_method	how cells were called from barcodes (e.g. emptyDrops)
median_genes	median genes detected per cell
median_umis	median UMIs (transcript counts) per cell
median_percent_mt	median percentage of mitochondrial reads per cell

A low median_genes or a high median_percent_mt relative to the other samples points to stressed or dying cells in that sample.