GapCloser from SOAP

[christinawu2008]

Hi, I just tried GapCloser from SOAP and it seems breaking some scaffolds down which causes a decrease of N50 length... what happens?

It indeed closes a large part of gaps (Ns from 31.5% to 4.4%), but the resulting scaffolds have a smaller total size and N50....

Christina

[Wallysb01]

I saw this too and I think it was actually a good thing. After doing my best to evaluate some of the scaffold that shrunk, they often did so because of faulty matepair reads, which were actually paired end reads. Also, it seemed to often be a case where the estimated insert size on the reads making the scaffold was just too big, and through gap closing, that was corrected.

You should definitely try to evaluate whats going on in your own data. Hopefully you have some closely related species you could map each long/short version of the scaffold to (I found blat on UCSC worked well for this purpose).

[christinawu2008]

I saw this too and I think it was actually a good thing. After doing my best to evaluate some of the scaffold that shrunk, they often did so because of faulty matepair reads, which were actually paired end reads. Also, it seemed to often be a case where the estimated insert size on the reads making the scaffold was just too big, and through gap closing, that was corrected.

You should definitely try to evaluate whats going on in your own data. Hopefully you have some closely related species you could map each long/short version of the scaffold to (I found blat on UCSC worked well for this purpose).

Thanks. I haven't employed mate pairs. It's right that the gaps are sometimes estimated large because the insert size of pairs are normally following a distribution, which we just set a constant value for scaffolding...