tophat/cufflinks bam vs. RPKM

[mgogol]

Is there any legitimate reason a tophat bam file would have reads for a gene whose RPKM was not reported in the output from tophat or cufflinks?

[thinkRNA]

Is there any legitimate reason a tophat bam file would have reads for a gene whose RPKM was not reported in the output from tophat or cufflinks?

You are finding reads in tophat bam file that are not reported in output from tophat (which is .bam format)? I don't really understand your question-can you give an example.
I thought tophat uses FPKM units and not RKPM.

[mgogol]

If I look at the list of genes whose FPKMs are reported by cufflinks, a certain group of genes is not found in that list at all.

If I load a bam file generated from the tophat sam file into IGV, and browse around looking at those genes, they have reads. So they have reads, but FPKMs are not reported.

I'm just wondering if this an error or if there is some legitimate explanation?

[thinkRNA]

if the reads are too few, they may be filtered by the F/--min-isoform-fraction parameter where TopHat filters out junctions supported by too few alignments.
May be you can set it to 0.

Also, I don't intuitively understand -a/--min-anchor-length <int> parameter, but it may also be limiting your reads that are counted for the FPKM.

[mgogol]

Thanks for the ideas.

[thinkRNA]

sure, if you find out whats going on, let us know. If the reads you saw on IGV don't span junctions, then its probably a parameter in bowtie that is throwing it- off-low quality or reads mapping to multiple places in the genome or something else. Its a good thing you caught it though