This is a mouse liver single-cell data comprising of 4 different experiments with 3 different protocols.
scMerge-integrated, this trajectory was estimated with higher degree of biological interpretability.
|Name||ID||Author||DOI or URL||Protocol||Organism||Tissue||# of cell types||# of cells||# of batches|
scMergearticle: Main Figure 3b and Supplementary Figure 8.
To combine the four liver scRNA-seq datasets, we used a semi-supervised approach by considering the developmental stages as the wanted variation to identify the pseudo-replicates. Note that the cells of the four liver datasets are from different fetal mouse liver developmental stages (E9.5-E17.5). To identify the pseudo-replicates corresponding to the hepatoblasts, hepatocytes and cholangiocytes, we used three known markers of hepatoblasts and cholangiocytes, Alb, Afp and Epcam, to guide the
scMerge algorithm. The sets of pseudo-replicates that are highly expressed the markers are further split according to the developmental stages.
In addition to integrating 4 scRNA-Seq datasets,
scMerge revealed a novel result: We constructed the cell trajectories with cells corresponding to the E17.5 time point of GSE90047 removed. We found that the trajectory associated with
scMerge is most consistent with the full Liver data collection and agrees with current literature, while other methods tended to generate extraneous branches with the subset of the Liver data collection.
We further performed SC3 on the
scMerge integrated data (k=9). We found that the SC3 clustering results have high concordance with the original cell types.
Data availability: Mouse Liver Data (in RData format)
scMerge parameters for integration: