Question: A few negative control probes (NCPs) are high in my Xenium data. Is the data still usable?
Answer: Data can be usable. We recommend careful consideration. The article What is the Xenium ‘Negative control probe counts per control per cell’ metric? explains Xenium negative control probes.
In the run summary report Decoding tab, examine the Gene-Specific Transcript Quality plot. If (i) the body of the majority of NCPs do not overlap the genes, (ii) a few NCPs trigger the high NCP alert, and (iii) the same tissue type consistently shows the same NCPs are high, then the hypothesis is a few probes are specifically binding and undergoing ligation and amplification to RCPs. For example, researchers have reported elevation of NegControlProbe_00022 and NegControlProbe_00017 in human brain panel samples. In such cases, we recommend researchers continue to use the sample data.
To confirm a few NCPs are driving the high NCP alert, recalculate negative control probe metrics using the transcripts matrix counts. Alternatively, run xeniumranger relabel with a modified gene panel JSON to exclude the outlier NCPs in the summary metrics. Xenium run results recapitulate the gene_panel.json. Given the aim of checking metrics, discard results afterwards and continue downstream analyses with original run data.
Step 1: Deprecate each NCP in the panel JSON
Before gene_panel.json
After modified-panel.json
Step 2: Run xeniumranger relabel with the modified panel JSON
xeniumranger relabel \
--xenium-bundle path-to-bundle \
--panel modified-panel.json \
--id new-results \
--localcores 8 --localmem 64
Review the web summary report for whether the high NCP alert is still present and for updated Negative control probe counts per control per cell metric values. The absence of the alert confirms the few NCPs contributed to the high rate. For samples where the alert persists, consider whether the sample has lower transcript density in the cohort, e.g. lower Total high quality decoded transcripts or Median transcripts per cell and further vet the quality of data against a truth set. Compare the sample to single-cell or other samples that represent the expected transcriptional profile, e.g. in a pseudo-bulk plot that uses the sum of transcript counts per gene. Consider using the single-cell data that informed Xenium panel selection or design.
If the high NCP alert is still present or if more than a few NCPs overlap the genes in the Gene-Specific Transcript Quality plot to begin with, then review sample preparation workflow. Contact support@10xgenomics.com and attach Xenium web summary reports to discuss optimizing sample preparation after considering the following.
- Regardless of negative control probe counts, samples that present lower transcripts per cell are concerning.
- Spatiality can help interpret the negative control rate further. Do the high quality (>= Q20) negative controls display spatiality, i.e. are more concentrated in non-tissue regions or are evenly dispersed across the tissue? If negative controls are concentrated in non-tissue areas or specific morphological regions, then factor for this when interpreting data quality. Some illustrative code towards plotting Xenium data is at https://kb.10xgenomics.com/hc/en-us/articles/11636252598925.
Related content
- What is the Xenium ‘Negative control probe counts per control per cell’ metric?
- The adjusted negative control probe rate appears elevated, what does this mean?
Product: Xenium In Situ Gene Expression
Last modified: February 15, 2024