StringTie software?

[amit491]

hi everyone,
I figured out the stupidity and that was -not reading the manual properly :-(

The Manual page says -

..If you run StringTie, and all FPKM values that were produced appear to have the value inf then it is very likely that the HI tag starts at 1. An easy command that you can run...

http://ccb.jhu.edu/software/stringtie/

So STAR out BAM files have the HI tag starting from 1 and hence.

[amit491]

Also to add, for the same BAM (STAR returned) on which Cufflinks successfully returned FPKM values, StringTie is giving me "Inf" and "-nan"
:-(

[amit491]

Hi folks,
Sorry for bumping an old thread but I am facing this weird output from StringTie and thought you could spot the issue.
I am running StringTie on STAR returned BAM files (I have used the option to insert XS tag in SAM/BAM fields)

I have now run StringTie on 6 samples and in each I have the same scenario -
FPKM value as "inf" or "-nan"

StringTie returns a GTF file and this an example of the 9th col for one transcript -

gene_id "STRG.67"; transcript_id "STRG.67.1"; reference_id "ENST00000447513"; ref_gene_id "ENSG00000157911"; ref_gene_name "PEX10"; cov "1.525000"; FPKM "inf";

About 98% of the transcripts have FPKM as "inf" and rest fraction as "-nan"

This is the cmd I am using

/path/to/stringtie-1.0.4/stringtie \
Aligned.sortedByCoord.out.bam \
-o StringTie_out/StringTie.gtf \
-p 16 \
-G /path/to/Homo_sapiens.GRCh37.73_Only-REF.gtf \
-B \
-e

I have checked the STAR BAM files and they seem ok: >85% reads with unique alignment.

Any hunches? I am sure I am doing something stupid but can't figure out what.

[colindaven]

Bayesembler is nice, and also works with large quantities of data rapidly.

I haven't successfully tried Stringtie yet, think an older version aborted with errors.

[robp]

It looks pretty good from the paper, and I'd definitely give StringTie a shot, but I've not tested the software yet. However, if you're looking to try out a new tool, I'd also take a look at the Bayesembler (http://genomebiology.com/2014/15/10/501). It seems to have a very nice and principled probabilistic model that it uses to jointly infer what transcripts are present and quantify their abundance.

[jake13]

I saw this in the latest issue of nature biotech: http://www.nature.com/nbt/journal/va.../nbt.3122.html. Here is the site for it: http://ccb.jhu.edu/software/stringtie/. Has anyone used this on their data yet? If so what did you think? I'm trying to estimate isoform abundance in my RNA-seq data.