Cufflinks - fragment mean length for single-end reads

[slavailn]

Thanks for reply, Jon!

The program runs with no errors and produces expected output files, but I'm a bit worried about Read Type shown as 0 bp single end? While it should be 62 bp. Also I'm not sure if I can leave Mean and Std Dev at defaults, considering that I had 62bp + 7 barcode + adapter. Would this influence the results?

Upper Quartile: 252.00
> Number of Multi-Reads: 508671 (with 1031383 total hits)
> Read Type: 0bp single-end
> Fragment Length Distribution: Truncated Gaussian (default)
> Default Mean: 200
> Default Std Dev: 80
[14:36:12] Modeling fragment count overdispersion.
[14:36:13] Calculating initial abundance estimates for multi-read correction.
> Processed 31353 loci. [*************************] 100%
[14:55:46] Testing for differential expression and regulation in locus.
> Processed 31353 loci. [*************************] 100%

[Jon_Keats]

This is a bit of a guess but my best guess is to leave these blank (ie. don't put in the command) for single end reads. Check the cufflinks output on stdout, I'm betting if figures out these are single end reads, and runs as such were those two options are not used.

[slavailn]

Hi, Mark!
I'm facing same kind of problem with single-end reads. Have you been able to resolve this issue?

Thanks!
Slava.

[marcho]

Hi,

so this is probably going to be a very naive question, but here goes anyway:

I have just started using cufflinks - the dataset is single-end reads from the HiSeq2000 platform with a length of 100bp (105 if you count the adapter, which has already been removed).

I noticed a dramatic difference in FPKM value estimates between the 'default settings' for fragment length man (-m) and standard length deviation (-s). Apparently, cufflinks is not able to figure out the correct values from the data. But what has to go in there, exactly? And what program would be best suited to give me those estimates? I used the Picard toolbox to get mean fragment length (100, surprise surprise) - but standard deviation is not determined there.

Also, while I would guess that 100 is the value to put in, I used -q 15 in the bwa assembly, which performed quite a bit if soft clipping on some reads. Does that factor in in any way?

Thanks for any helpful comments!

/Marc