Begin your QC analysis with the library-level metrics, as these sequencing and mapping metrics are foundational for assessing library quality. If these metrics show issues, it likely indicates compromised quality across the library and potentially poor sample-level metrics (multiplexed experiments). First, verify the number of barcodes assigned as cells (cell calling). Start by checking the estimated number of cells metric to ensure that it aligns with your expectations.

Use the barcode rank plot to confirm the expected number of cell-associated barcodes and to check for single cell behavior. An unusual shape in the barcode rank plot at the library level may suggest issues with GEM well loading or a potential wetting failure. Refer to the sample-level barcode rank plot to confirm single cell behavior for this sample (described in the sample tab).

Next, look at the fraction of reads in cells metric. A low fraction of reads in cells confirms the absence of single cell behavior. Since this is a library-level metric, check all the samples in your experiment to identify any problematic sample. Excluding problematic samples from downstream analysis may improve results.

Lastly, the fraction of initial cell barcodes passing high occupancy GEM filtering metric helps verify proper chip loading. A high value here indicates successful chip loading, while lower values suggest data loss, which may require cautious downstream analysis.

Use the mapping metrics to assess whether your reads mapped well to the reference. Each metric in the mapping quality section provides insights into mapping quality.

• Reads mapped to the probe set: Lower values may indicate low total expression, use of incorrect probe set, suboptimal sample preparation, or the use of input FASTQs from the wrong assays (i.e., non-Flex).

• Reads confidently mapped to the probe set: This fraction should also be high. If reads mapped to the probe set are high but reads confidently mapped are low, run FastQC to troubleshoot.

• Reads half-mapped to probe set and reads split-mapped to probe set: Evaluate these metrics to check for partial mappings. Higher values for these metrics may result from low RNA content in the sample, inadequate washing after probe hybridization, deviations from the recommended protocol during hybridization, or suboptimal sample preparation.

Check the sequencing depth to ensure sufficient total reads and reads per cell. If these metrics look satisfactory, proceed to the following metrics to confirm read quality:

• Valid Barcodes
• Valid UMIs
• Valid Probe Barcodes
• Valid GEM barcodes

If you observe low values for the above four metrics, check the Q30 metrics and consult your sequencing provider for further guidance.

For Flex experiments, this section helps assess whether the distribution of UMIs and cells among samples is as expected.

• UMIs per probe barcode: Check the number and fraction of UMIs for each Probe Barcode among all UMIs for that library type. The fraction of the UMIs should be similar between probe barcodes if the associated samples are expected to have similar complexity and target cell numbers.

• Cells per probe barcode: Similarly, the number and fraction of cells per Probe Barcode among all cells detected in the GEM well should be similar if the associated samples are expected to have similar complexity and target cell numbers.

Will sequencing deeper help?

Use the sequencing saturation plot and median genes per cell plot to assess if increased sequencing depth could improve your results. If the slope of the median genes per cell curve near the endpoint is flat, it suggests saturation, indicating that more sequencing depth is unlikely to increase transcript complexity.

Is there genomic DNA noise?

The estimated UMIs from genomic DNA metric should be less than 1% under optimal experimental conditions. Similarly, the estimated UMIs from genomic DNA per unspliced probe metric should be low.

Evaluating transcript complexity

For each sample, examine the median genes per cell plot on its web summary to check if there is sufficient transcript complexity. The curve's slope near the endpoint offers an estimate of the potential gain from increasing sequencing depth. A flat slope indicates saturation, suggesting that additional sequencing will not enhance transcript complexity.

If you observe distinct structures or unexpected patterns in these visualizations, contact 10x Genomics Technical Support for further troubleshooting or filter the data and proceed with caution.

Checking expression of specific genes

To see if specific genes of interest are highly expressed in your cells:

• Examine the t-SNE projection of cells colored by cluster to identify clusters of interest.
• Review the Top Features By Cluster table for key gene expression patterns.

The Cells count reported in this section matches the library-level cell calls. Check the Fraction antibody reads to confirm that this value is high. If the percentage is low (less than 70%), verify that the sections in your Feature Reference CSV were correctly specified. Common issues include incorrect entries in the sequence or pattern fields. Additionally, low-quality reads may contribute to a lower fraction of antibody reads.

The Antibody reads in cells section reports the fraction of valid-barcode, valid-UMI, recognized antibody Feature Barcode reads associated with cell barcodes. This metric should also be high. Low values can result from various factors, including:

• Poor cell washing
• Suboptimal sample quality
• Emulsion failures
• Suboptimal antibody titration and labeling methods
• Low target population (e.g., use of rare markers)
• Low sequencing depth, which can impact cell-calling algorithms

The Fraction of antibody reads usable should ideally exceed 20%. This metric can be affected by a low number of antibody reads in cells and a low fraction of antibody reads.

The Fraction of antibody reads in aggregate barcodes metric shows the fraction of read pairs with valid barcodes that were removed due to aggregation. Ideally, this value should be below 5%. Higher values may result from poor sample quality, inadequate washes, or suboptimal staining and antibody titration. Pre-spinning antibodies is also essential to remove protein aggregates.

Evaluate the sequencing quality of the Antibody Capture library in a similar way to the Gene Expression library. A key metric to examine is Mean reads per cell, which is calculated as the total number of sequenced antibody read pairs divided by the number of cell-associated barcodes.

The Metrics per probe barcode section can also be evaluated similarly to the Gene Expression section. The UMIs per probe barcode column reflects UMIs specifically from antibody reads. The Cells per probe barcode column pertains to the entire Gene Expression library and should match the values reported in the Gene Expression library tab.

On the Sample tab, double-check the sample ID. In a multiplex experiment, there will be one web summary per sample. Sample tabs are not present for singleplex experiments.

The Sample tab includes one or more subtabs based on your experimental design, such as Gene Expression, Antibody (or other Feature Barcode library), and Cell Annotation (if cell annotations were enabled in the multi run).