Prokaryotic RNA-seq / Mixing IonTorrent and Illumina data

Question

[caramba]

I have sequenced RNA from a bacterial culture grown under different conditions. The control sample was sequenced with IonTorrent, the other three other treatments were sequenced with Illumina HiSeq. Both samples were rRNA depleted before sequencing. I'd like to use TMM-normalized CPM to compare expression of different genes of interested due to not having replicates. Is that ok in principle or would there be a more suitable approach?

I noticed that many reads from IonTorrent data map to small RNA (5S, tRNA) which isn't the case for the Illumina data. As far as I understand, CPM normalizes by the total read count that mapped and would lead to a significant bias between IonTorrent and Illumina dat, correct? Any advice on how I could deal with this?

Thanks.

Answer 1

[Carlo Yague]

Comparing RNA-seq results from two different platforms is not the ideal case at all but I suppose you don't have a choice here.

A possible approach would be to compute read counts per genes using HTseq-counts or featureCounts omitting features shorter than a reasonable treshold (like 500 bp), then to normalize according to that distribution (DESeq does that with quantile normalization, edgeR with TMM-normalization).