Differential Expression and Co-expression analysis from Cell type-specific datasets

Question

[volvicpellegrino]

I use translating ribosome affinity purification (TRAP - the same principle as RiboTag) to extract translating mRNA from a specific cell type, where ribosomes are GFP-tagged. I have an RNA-seq dataset from TRAP-processed samples, that are enriched in genetic markers specific to this cell type. I want to study the effects of different conditions on this cell type through differential gene expression (DESeq2) and co-expression analysis (WGCNA).

Using DESeq2 on kallisto counts: Out of ~23,000 genes, ~4000 are significantly enriched in my TRAP RNA compared to RNA from the whole tissue input (Total RNA). ~5000 genes are significantly depleted in TRAP RNA compared to Total RNA (these include specific markers of non-target cell types).

My question: Should I filter genes before performing TRAP sample vs TRAP sample comparisons under different conditions, to remove the 5000 depleted genes?

A paper discussing analysis of TRAP datasets talks about filtering microarray probes with intensities lower than that of specific markers of non-target cell types. Would this make sense in an RNAseq scenario?

Answer 1

[benformatics]

The thing is that just because those genes aren't being actively translated (i.e. present in your TRAP data) doesn't mean they aren't present/transcribed at biologically significant levels in the cells.

TRAP sounds like an enrichment method of known targets so it make sense to use the true negatives to set a threshold for expression (as mentioned for the microarray analysis). In RNA-Seq you don't have that luxury.

-

You should however filter out "unexpressed" genes (e.g. genes below a certain expression threshold in the total RNA). This is standard protocol for RNA-Seq DGE (differential gene expression).

[volvicpellegrino]

Thanks for your reply, however I don't understand why the microarray threshold can't be used in RNA-Seq?

TRAP involves immunoprecipitating GFP tagged ribosomes that are conditionally expressed in a target cell type. There is some non-specific RNA from other cell types, such as astrocytes. I could use the abundance value for a cell marker (e.g. Gfap) as a threshold to filter genes with read counts below this?