Question:
Both Flex (probe-based) and 3’ GEX (reverse-transcription-based) assays are used to profile the whole transcriptome but with different chemistries. How does the data compare between the two assay types?
Answer:
Data from the Flex (i.e. Fixed RNA profiling or FRP) assay is highly comparable and correlated with the data from 3’ GEX assay. In fact, two of the probe validation criteria we used internally include:
- High pseudobulk correlation with the corresponding gene in the 3’ v3.1 GEX assay
- High correlation with the 3’ v3.1 GEX assay in the differential gene expression pattern over multiple cell types in a sample
Below are example comparisons showing the comparability between the two assays (point 1 - 4). For the datasets shown below, samples were analyzed using the Chromium Single Cell 3’ v3.1 assay (fresh), and were concurrently fixed and assessed using the Chromium Fixed RNA Profiling assay. Here we also provide some points to be considered when performing such data comparison (point 5 - 6).
1. Pseudobulk gene expression. The two example datasets here are from mouse PBMC and splenocytes, respectively. The reported R-squared value is calculated at matched depth (20k raw reads per cell) on the log-transformed UMI counts over all genes with non-zero UMI count in the 3' v3.1 GEX sample. This shows high correlation in the detected gene expression levels between Flex (FRP) and 3’ v3.1 GEX assay.
2. Sensitivity. The two example datasets here are from the same mouse PBMC and splenocyte samples mentioned in 1. At the same depth (reads per cell), the Flex assay has comparable or higher sensitivity compared to the 3’ v3.1 GEX assay.
3. Cell populations. The two example datasets here are from the same mouse PBMC and splenocyte samples mentioned in 1 and 2. The cell types and populations identified in the samples are similar between the two assay types.
4. Gene expression across cell populations. In the human breast cancer DTC example shown below, the gene expression pattern across different cell types is consistent between the Flex and 3’ v3.1 GEX assay. The two datasets, integrated by the Harmony package, had overlapping UMAPs indicating comparable sensitivity and specificity.
5. Batch effects. When comparing Flex data with 3’ v3.1 (or in general, reverse transcription-based) data, because the chemistries are fundamentally different (probe-based v.s. reverse transcription-based), significant batch effects are expected when combining the two types of data. Therefore, it is likely that the clusters from Flex v.s. 3’ v3.1 data do not overlap without batch correction. Moreover, direct comparison of raw count values between the two assays is not recommended due to the distinct difference in the chemistries. Normalization and batch correction should be performed before integrating and comparing the datasets. Please see this article for information on integration of Flex data with 3' or 5' GEX data.
6. Potential difference in transcriptional profiles observed between the assays. When looking at the gene expression in Flex v.s. reverse-transcription based assays, a few factors could be contributing to differences in the profiles:
- Some genes are not included in the probe set as mentioned on this page (Frequently asked questions, 1). In addition, some genes may be excluded from the probe set and analysis due to inability to design probes that meet our probe design and quality control criteria or the designed probes having predicted off-target activity. Please check the probe set csv file (download here) to confirm if the genes of interest are present and included in the analysis.
- Probe panels were designed to dampen signals from some genes (e.g. highly expressed genes and mitochondrial protein coding genes) as mentioned on this page (Frequently asked questions, 2). Therefore, the relative signal from these genes may seem lower as compared to what one would normally expect from a reverse transcription-based assay. You could see the probe set metadata TSV file (download here) to check the coverage of each gene.
- There may be a boost in the sensitivity for genes that are expressed at low levels and may not normally be detected as well in reverse-transcription based chemistries.
Last updated: July 2023