Skip to main content
. 2019 Apr 26;35(22):4688–4695. doi: 10.1093/bioinformatics/btz292

Table 3.

Annotation of 93 655 PBMC cells profiled on the 10X platform

Specific genes
Highly expressed genes
All genes
Pearson (%) Spearman (%) Pearson (%) Spearman (%) Pearson (%) Spearman (%)
CD4+/CD25+ regulatory T cells 37.7 8.4 55.6 53.2 59.7 19.3
CD8+ cytotoxic T cells 30.7 37.5 2.0 76.4 2.0 98.3
CD19+ B cells 92.0 41.5 38.1 99.9 37.4 100.0
CD56+ NK cells 0.6 4.9 22.9 0.0 20.6 0.0
CD4+ helper T cells 55.5 68.1 98.2 30.0 98.4 5.8
CD14+ monocytes 10.7 78.8 28.0 42.9 27.5 78.8
CD34+ cells 0.8 4.4 56.3 83.0 36.0 82.8
CD4+/CD45RO+ memory T cells 44.0 0.5 26.3 12.7 28.5 5.4
CD4+/CD45RA+/CD25− naive T cells 11.1 51.0 86.6 5.5 85.9 7.7
CD8+/CD45RA+ naive cytotoxic T cells 29.7 35.8 0.0 86.5 0.0 99.1

Note: The relative recalls for PBMCs when annotated using scMatch with the FANTOM5 dataset as a reference. (i) Pearson’s correlation coefficient and (ii) Spearman’s correlation. Accuracy calculated using different gene lists is shown [(i) All reference set genes (22 049 genes), (ii) Highly expressed genes (detected in the FANTOM5 atlas with maximum expression ≥500 tags per million, 4129 genes) and (iii) Manually curated lineage-specific genes from the FANTOM5 atlas (272 genes)]. Annotation recall is calculated as the number of cells that were correctly annotated by scMatch divided by the total number of cells of that type in the Zheng et al. (2017) dataset.