thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Single-Cell Analysis questions

Question 14 of 20 Intermediate Troubleshooting

A Cell Ranger summary shows a low fraction of reads in cells. How would you investigate?

Short answer

A low fraction of reads in cells usually means much of the RNA was ambient, free in the suspension and captured by every droplet. Check viability and the barcode rank plot first, then whether cell calling missed a population with low RNA content.

Why

Fraction reads in cells is the share of reads with a valid barcode, mapped confidently to the transcriptome, whose barcode was called as a cell. The 10x Genomics technical note on web summaries gives over 70 percent as the expected value and says lower percentages indicate a high level of ambient RNA partitioned into all droplets. Ambient RNA comes from dead and lysed cells, so the usual root cause is a stressed or dying suspension, or harsh dissociation.

The barcode rank plot helps separate causes. A round curve with no steep cliff suggests low viability, premature lysis or a wetting failure. A clean cliff with fewer cells than expected suggests a clog or a wrong cell count.

A plot with two cliffs means two populations with very different RNA content, and cell calling may have missed the smaller cells, which leaves their reads outside the called cells. The consequences reach the biology. High ambient RNA spreads marker genes from abundant cell types into every cell and can blur annotation.

How to work it through

  1. Read the fraction reads in cells alongside the estimated cell number and median genes per cell.
  2. Inspect the barcode rank plot for a steep cliff, a round curve or two cliffs.
  3. Check the lab record for viability, dissociation time and how the cell count was made.
  4. If two populations are suspected, rerun cell calling on the raw matrix and see whether low-RNA cells are recovered.
  5. Estimate ambient contamination and check whether marker genes of abundant cell types appear in unrelated clusters.
  6. Decide whether to correct for ambient RNA, accept the data with caveats or repeat the sample.

An example

A tumor sample shows 48 percent reads in cells, 3,100 estimated cells against 8,000 targeted, and a barcode rank plot with a shallow second shoulder. The lab notes 62 percent viability after a 90 minute digestion. Rerunning cell calling on the raw matrix recovers 1,400 additional barcodes with low RNA content that cluster as neutrophils.

Hemoglobin and immunoglobulin genes appear across all clusters, a sign of ambient spread. The team corrects for ambient RNA, keeps the neutrophils, and shortens digestion for the next samples.

Where the answer stops holding

The 70 percent figure is a general guide from one technical note and will vary by tissue and chemistry. Ambient correction methods estimate contamination and can over-correct genuinely shared genes. Some tissues, such as those rich in small, low-RNA cells, will always sit lower without the data being unusable.

A strong answer covers

  • Low fraction reads in cells points to ambient RNA
  • 10x expects over 70 percent in its technical note
  • The barcode rank plot separates viability, clog and mixed-population causes
  • Low-RNA cells may have been missed by cell calling
  • Ambient RNA spreads abundant markers into all cells

Common mistakes

  • Resequencing deeper, which adds reads but does not change the fraction in cells
  • Ignoring a second cliff and losing a real low-RNA population
  • Annotating clusters without checking for ambient marker spread