Successive rounds of gapfilling and scaffolding

[Zapp]

Hi,

we've been playing around with successive rounds of gapfilling and scaffolding and we see a constant improvement of our genome stats. Basically we run several iterations of gapfilling which generates new ends with potential mapping sites for our mate-pair libraries and re-scaffold the genome. However, besides not being able to really assess the quality of these so generated new scaffolds, we wondered if re-scaffolding after gapfilling should be performed on contig level i.e. gapfilling, breaking down the scaffolds to contigs (now longer thanks to the gapfilling) and rescaffolding, or if it would be rather advisable to re-scaffold the gapfilled scaffolds. Theoretically, breaking down the scaffolds to contigs after gapfilling would allow to intercalate contigs that could not be mapped to the previously existing gaps due to the missing sequence at the end of the contigs before gap filling.

Has anyone played around with this and can share his experience?

Thanks,
Zapp

[Wallysb01]

Gapfilled regions of your genome will be of lower quality than the parts created by kmer extension in the original assembly, so too many rounds of gapfilling will start to introduce more error. Because of this, I would lean towards keeping your scaffolds intact for further rounds of scaffolding. This way any new mismapping mate pairs in gap filled regions don’t invalidate and screw up what should be higher quality scaffolding links already made.

[Zapp]

That sound reasonable, thank you very much.

Regards,
Zapp

[flxlex]

Stand-alone scaffolding will never be as good a built-in scaffolder that can take in all information used for the assembly...