LieberInstitute/10xPilot_snRNAseq-human

Analysis saved in SingleCellExperiment object

Closed this issue · 3 comments

jhusarven commented

I downloaded the Lieber 10x single cell data in order to determine what
cell types express my genes of interest (I have a list of ~200 genes).
I am new to single cell data and singlecellexperiment objects and I just
want to use the published/existing analysis.

I used rowData(sce.dlpfc.tran) to get a table that has column names like
propExprsIn.Astro, propExprsIn.Excit_A... These appear to be normalized
data (0,1) and I was planning on using these values for determining the
cell types that express my genes of interest. I wanted to make sure that this
is correct. If not, what should I use?

Thanks in advance.

Best,

Sarven

mattntran commented

Hi Sarven,

Thanks for the interest in our published datasets!

Those are actually not the normalized count data; it refers to the proportion of cells in the respective j cluster that express any amount of gene i (i.e. deemed by non-0 expression in this case). This was meant to provide a quick glance as to where a gene might be highly expressed. However, we would argue this is only partially informative, as it doesn't give insight as to how much of a gene is expressed, e.g. the relative log2 counts of gene i in cluster j, compared to cluster k, etc.

If you were interested in the normalized counts, you'll want to work with assay(sce.dlpfc.tran, "logcounts"). If you perhaps want to cross reference your ~200 genes with cell class markers beyond the top 40 we provide in this section, it could be useful to identify cluster markers following our (or a similar) approach, then identify the intersecting genes.

And of course a more interactive (though laborious in its own regard) approach is to explore your genes of interest in our shiny apps, for the respective regions of interest. One feature I may point out is the complex heatmap panel: there you can click on 'Data parameters' > 'Edit feature names', which pulls up a text editor where you can then list your genes (one gene per row), to observe any potential cluster specificity or enrichment.

Hope this gives some helpful direction!

Cheers,
Matt

lcolladotor commented

I'm closing this issue due to lack of activity + it seems like Matt already resolved it. Sarven, feel free to re-open the issue if needed.

jhusarven commented