Five scRNA-seq datasets from human liver cells (MacParland et al, 2018), mouse retinal neurons (Shekhar et al, 2016), Tabula Muris (Tabula Muris Consortium, 2018) and two peripheral blood mononuclear cell (PBMCs) datasets (Zheng et al, 2017) and (Gierahn et al., 2017) were processed and curated as described by Diaz-Mejia et al 2019 to use them as inputs for benchmarking of cell cluster labeling methods. Five types of files are provided: a) Single-cell RNA-sequencing (scRNA-seq) cell cluster average gene expression matrices (*E_xy_matrix.tsv) b) Cell type gene expression signatures in the form of gene sets (*gmt), binary (*binary_profile.tsv) and continuous profiles (*continuous_profile.tsv) c) Reference cell type gold standard annotations (*gold_standards.tsv) d) Supplementary Table 1, containing ROC AUC, PR AUC and computing times reported by Diaz-Mejia et al 2019 in Figure 6. e) Supplementary Table 2, containing ROC AUC, PR AUC analysis of the PBMC datasets using either the LM22 (Newman et al 2015) or the Monaco et al (2019) cell type signatures. Note: as described in Diaz-Mejia et al 2019, the PBMC and Tabula Muris datasets were analyzed in two ways each: i) pbmc-22-10x and pbmc-6-10x, ii) pbmc-22-seqwell and pbmc-6-seqwell, and iii) tabula_muris_11 and tabula_muris_6.

Version 2.1.1 of this dataset corresponds to the Version 3 of the Diaz-Mejia et at., (2019) F1000 Research publication. Main differences compared to v2.1.0: - Added Supplementary Table 2 - Added percentage of clusters correctly assigned to Supplementary Table 1

{"references": ["Diaz-Mejia et al (2019). https://doi.org/10.12688/f1000research.18490.1"]}