mate pair insert size variation and de novo assembly

[Mark]

Hi All

One of my lab bench colleagues is working on a constructing an Illumina mate pair library intended for use in a target capture procedure where a substantial amount of DNA is needed. Unfortunately, if she does size selection on this library she loses 90+% of her library and so she is wondering what the consequences of not doing the size selection will have on de novo assembly using this library. I believe she is attempting to make a 6kb insert library but but the size of the fragments varies from ~3-10kb. I would probably try to assemble this data (with short insert paired end data as well) using velvet but don't have much experience with mate pair libraries. Can anyone comment as to how detrimental this level of insert size variation would be in a velvet assembly or make suggestions how to deal with it?

Thanks

Mark

[Torst]

That fragment range doesn't sound too bad. Even AFTER the protocol, I see size ranges that wide, and successfully assembled them.

[Wallysb01]

After a certain point you just won't care how many Ns are in gaps. So if you're scaffolding to a mean of 5000 bp insert, you might be wrong by 2kbp, but do you care?

Unless you have the intention of ever filling such large gaps, this really shouldn't matter for you. And if you do try to fill them, you'll be using tools that understand you might have +/- 50% error in the lengths gaps of that size, and thus they will rely on data from other sources to fix it.