CuffDiff: genes with highly abundant reads have FPKM 0.

[uniportdb]

I just started learning rna-seq data anlaysis following Nature Protocols 7, 562–578 (2012).
I used CuffDiff after tophat, but the genes with highly abundant reads have FPKM 0.
I tried all the different normalization options in CuffDiff, and --max-bundle-frags 2000000, but still FPKM 0 for the genes with the largest number of mapped reads. Is there any upper bound like Microarray excludes the saturated spots ?

I added example below,
In sample A, it has FPKM 0, because it has the most number of reads, but the same gene in sample B has FPKM 3789, because it's in the confidence zone in sample B ? Please, help me to get FPKM for those genes with highly abundant reads. How can I get FPKM for PM_0405A60 from the example below?
Thanks,


Both examples below were sorted by the # reads in the descending order.
I counted the number of reads using 'htseq-count'

Sample A

Gene | # reads | FPKM
PYYM_0405060 | 1447079 | 0
PYYM_1351050 | 675926 | 0
PYYM_1007060 | 559162 | 0
PYYM_1209040 | 421148 | 29315.7

Sample: B
Gene | # reads | FPKM
PYYM_0501060 | 1633262 | 0
PYYM_1351050 | 757339 | 0
PYYM_0405060 | 552553 | 3789.85

CuffDiff –verbose printed the below; PM_0405A60’s loci is PyYM_04_v1:214849-216007


Inspecting bundle PyYM_04_v1:206968-210685 with 5956 reads
Inspecting bundle PyYM_04_v1:211950-212458 with 368 reads
Inspecting bundle PyYM_04_v1:213051-213787 with 2864 reads

Inspecting bundle PyYM_04_v1:214849-216007 with 943663 reads

Inspecting bundle PyYM_04_v1:218737-223186 with 746 reads
Inspecting bundle PyYM_04_v1:223748-224267 with 889 reads
Inspecting bundle PyYM_04_v1:225620-226957 with 399 reads
Inspecting bundle PyYM_04_v1:227958-238759 with 10900 reads

[paulk]

I encountered something like this last year, where I was getting transcripts flagged as HIDATA in Cufflinks output.

You might try increasing the --max-bundle-frags setting in from its default setting to get around this issue. See:

--max-bundle-frags <int> Sets the maximum number of fragments a locus may have before being skipped. Skipped loci are listed in skipped.gtf. Default: 1000000

http://cufflinks.cbcb.umd.edu/manual.html

[kadircaner]

I'm having the same problem but in my case there are no genes in the skipped.gtf file, so it's not because of HIDATA flag but there is another problem that I couldnt figure out yet... Any help would be really appreciated!!

[uniportdb]

Thank you, PaulK.

I tried --max-bundle-frags 2000000, but still 0 FPKM.
And then I tried --max-bundle-frags 3000000, at last I have none-zero FPKM 107219.
I guess it's paired-end, so the max should be > 2 * No. reads.