I am currently working with single-cell RNA-seq data generated using the Parse Biosciences WT Mega Kit. We also have data generated using 10x Genomics from similar types of samples, allowing us to compare the two technologies.

I am observing a substantial difference in the clustering and cell-type separation between the two datasets.

In the 10x dataset, the different cell populations are relatively well separated in UMAP space. Moreover, when I examine the top differentially expressed genes (DEGs) for each cluster, they include well-known, cell-type-specific marker genes, which makes cell-type annotation relatively straightforward.

In contrast, in the Parse dataset, the clusters are much less distinct. Most of the clusters appear to be grouped together and form more of a continuous structure rather than clearly separated populations (I have attached an example UMAP for reference). More importantly, when I examine the top five DEGs for the clusters, they do not show the expected cell-type specificity, making cell-type annotation considerably more difficult.

I would therefore like to understand whether this could be related to the characteristics of Parse Biosciences data or whether there are specific preprocessing, normalization, feature-selection, dimensionality-reduction, or clustering steps that need to be adjusted for Parse data.

Has anyone experienced similar difficulties with cell-type separation or marker-gene identification when analysing Parse Biosciences scRNA-seq data?

Any advice or experience with this issue would be greatly appreciated.

Thanks in advance.

Best regards,

Nitin N.