reference transcripts missing after cuffmerge

[jake13]

I ran cufflinks on my rna-seq samples and then ran cuffmerge with the -g (reference gtf) and -s (reference sequence) options, which I thought would merge all of the reference transcripts with my cufflinks files. However, when I run cuffcompare on this merged gtf vs the reference gtf, the merged gtf is missing a number of transcripts. Does anyone know why?
Thanks

[yueluo]

What do you mean by 'the merged gtf missing a number of transcripts' ?
Can you describe what you did in detail?
Specifically -- In what order did you run cufflinks/cuffmerge/cuffcompare, and with what options / input files for each process.

[jake13]

I first ran cufflinks on each sample with the following command:

cufflinks -p 16 -u -g $gencode_gtf -b $gencode_fasta -M $gencode_rRNA --library-type fr-firststrand sample1.bam

cufflinks -p 16 -u -g $gencode_gtf -b $gencode_fasta -M $gencode_rRNA --library-type fr-firststrand sample2.bam

This produced a transcripts.gtf file for sample1 and a transcripts.gtf file for sample2.

I then ran cuffmerge with the following command:

cuffmerge -g $gencode_gtf -s $gencode_fasta assembly-gtf.txt

where assembly-gtf.txt contained the path to sample1 and sample2 transcripts.gtf

I then got a merged.gtf file. I wanted to see how many novel transcripts I found between these two samples compared to the reference assembly so I ran cuffcompare with the following command:

cuffcompare -r $genocde_gtf -s $gencode_fasta merged.gtf

When I look at the stats it lists a number of missed exons and a number of missed loci.

[yueluo]

cuffmerge with -g <ref> merges transcripts from different samples, and then merges these transcipts with known transcripts. No reference transcript is included if it is not detected in any of your samples.