Hi all,
I'm interested in comparing the assemblies produced by Bayesembler and StringTie against a reference annotation using a few different metrics. In particular, I have reference alignments of a few RNA-seq samples in human (against hg19) and I'd like to see how many of the transcripts recovered by the different programs match known annotations well etc. Obviously, the different tools asses metrics like this in their respective manuscripts, but not on the same data set and not against each other. Is there a common / standard tool I can use to compare the different GTF files to get measurements like sensitivity and specificity without having to e.g. write my own evaluation pipeline? The biggest issue seems to be calling "accurate" transcripts that have only minor differences from the reference (e.g. a few bases off at the TSS or transcription termination site). For my purposes, I'd like such transcripts to be considered correct if they otherwise match the annotation. Any suggestions?
How did it turn out?