CollectRNASeqMetrics Picard - where do they get UTR regions?

[apredeus]

Hello all,

I've been using Picard CollectRNASeqMetrics tool for a while now - it is very useful for all kinds of RNA-seq metrics, as you probably know. It uses "refFlat" type of annotation file, which can be downloaded directly from UCSC table browser for RefSeq annotation, or obtained using gtfToGenePred utility for other annotations, such as Gencode.

Here's what I'm a bit puzzled about. It's clear how the program determines if a read lands on exon, intron, or an intergenic area. However it also provides and estimate for UTR read percentage. And this information is not in the annotation file. How does it know then - some sort of guessing algorithm?

Thank you in advance

[JamieHeather]

Upstream of ATG of first codon and downstream of last stop codon maybe?

[apredeus]

That's what I thought. It seems to overestimate that percentage quite dramatically. Maybe I can hack into their code so that it would use UTR annotation from Gencode...

[apredeus]

Actually, I was wrong - refflat format includes information about which part of the transcript is coding. So there's enough information to tell UTR from coding sequence.