Cuffdiff NOTEST for splicing.diff

[Greg]

Using the tophat/cufflinks combo to find differential alternative splicing.

The cuffdiff manual says that if there "not enough alignments for testing" then you will get NOTEST and the treatments will not be compared. Does anyone have an idea as to what the cutoff for this is? Many of the genes that get NOTEST in my data set (single end) have a large number of reads (several hundred). Could it be because of insufficient coverage? or maybe operator error?

thanks!

[Greg]

So I just found the -c option to set the number of required reads. It does change the number of tests that happen but there are still less then I would think. Any ideas? What are the other criteria for doing the test or not? Would more info on my workflow be helpful?

[griffon42]

I'm experiencing the same issue. I'm using cufflinks 0.8.3 (the 7/2 release) and the -c option doesn't seem to work as described in the manual. As extreme examples, I have loci that are "OK" for stat testing and "yes" for significant but have only 2 reads vs. 0 reads across samples.

As mentioned, changing -c DOES affect the number of loci tested, but not by much and it does not fix situations like that described above.

Does anyone know how to manage this situation? The differential expression output files i'm generating are rather unmanageable.

Thanks.

[jetspeeder]

I have a lot of NOTEST and FAIL as well,

Also do you guys know the significance of this and why the percentages are high for missed and wrong exons and what the signifies? Thanks.

#> Genomic sequence: chr1
# Query mRNAs : 20972 in 20055 loci (3356 multi-exon transcripts)
# (633 multi-transcript loci, ~1.0 transcripts per locus)
# Reference mRNAs : 1024 in 912 loci (966 multi-exon)
# Corresponding super-loci: 695
#--------------------| Sn | Sp | fSn | fSp
Base level: 63.4 16.4 - -
Exon level: 37.6 9.6 44.0 11.3
Intron level: 52.6 31.5 53.4 31.9
Intron chain level: 18.6 5.4 24.7 7.1
Transcript level: 0.0 0.0 0.1 0.0
Locus level: 19.5 0.9 22.9 1.0
Missed exons: 2554/8302 ( 30.8%)
Wrong exons: 26036/32327 ( 80.5%)
Missed introns: 3161/7289 ( 43.4%)
Wrong introns: 7825/12186 ( 64.2%)
Missed loci: 208/912 ( 22.8%)
Wrong loci: 17823/20055 ( 88.9%)

[Greg]

I've had the same experience. Could only guess...

[SEQond]

So anyone has an idea on how to find the ideal -c "minimum number of alignments in a locus for needed to conduct significance testing" (--min-alignment-count) ?