FPKM=0 in cufflinks - but there are reads mapped to these genes!

[dsher]

Hello everyone

We are using BAM to map Illumina reads to a bacterial genome, followed by Cufflinks to get the FPKMs (implemented in Galaxy). We have come across many genes for which we get FPKM=0 (using both gene and transcript expression) even though there are reads mapping to these gene IDs (when we visualize using IGV).

The genes which have"0 FPKM" have different tags: some are LOWDATA, some are HIDATA and some are OK...

We tried changing the number of reads used as input into Cufflinks - when we used many reads (~12,000,000 on a genome with ~2,400 genes) most of the genes have 0 FPKM and "HIDATA". When we used fewer reads (1,000,000) on the same genome many of the genes still had "0 FPKM" but now with different flags (OK or LOWDATA).

We tried using quartile normalization and this didn't seem to help much.

Why is Cufflinks saying there is no expression from expressed genes? Can anyone suggest a reason/fix for this?

Thanks
Daniel

[pbluescript]

From the Cufflinks manual:
--max-bundle-frags <int> Sets the maximum number of fragments a locus may have before being skipped. Skipped loci are marked with status HIDATA. Default: 1000000

Just make that option higher. You will need to be sure the high number of reads mapping there represent reality rather than some sort of artifact though.

[dsher]

Thanks for the tip - but this does not seem to be the case. We are using relatively few reads and get FPKM=0 with "OK" and not "HIDATA".

looking at IGV, it seems that in several of these cases the genes on the GF file overlap (this is a bacterial genome with high coding density). Could Cufflinks just skip these regions?