Picard::CollectRnaSeqMetrics Expecticed/Minimum Values for Sample QC

Question

[bkellman]

I'm using CollectRnaSeqMetrics from the Picard toolbox but I'm having trouble finding expected/minimum for the metrics. I would like to apply these metrics to QC RNAseq samples. There are some obvious issues (e.g. high rRNA%) but for other metrics it is less clear what is: concerning, problematic, acceptable...

Q1) Can I use these metrics to identify problematic/concerning samples?

Q2) Where can I find examples of good/okay/bad values for each of these metrics?

Thank you, Ben

Edit: We are using rRNA exclusion prep.

Answer 1

[fanli.gcb]

Broadly speaking, it depends on your library and sequencing strategy. For example:

• 3'->5' bias: If you are using an oligo(dT) strategy, then you may get overrepresentation of 3' ends. See http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3821180/ as an example
• Intron/exon content: Again, polyA+ libraries tend to have higher exon content, whereas ribo-depleted or other strategies may have more intronic content.

A few more useful papers:

• Technical Variations in Low-Input RNA-seq Methodologies
• Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells
• Evaluation of Commercially Available RNA Amplification Kits for RNA Sequencing Using Very Low Input Amounts of Total RNA

[bkellman]

Thanks, I'll check out the links. We used rRNA depletion for prep.