After genome assembly...

[cascoamarillo]

...what to do?
Hi, I've been working a little with a large genome assembly (250Mb). I have Illumina and PacBio reads using diffrent approaches. At the end, used SPAdes assembly with Illumina data and PBJelly for filling gaps with PacBio subreads. Decent assembly at the end.
But when I compare the assembly with some regions I have already sequenced, looks like the assembly can still be improved. Situations like this are found:

reference: ****************************************************************************************
contigs:    -------------------------------                              -----------------------------------
                                      -----------------------------------------------

where there are overlapping regions of 70-75 nt with 100% identity.
Is there any post assembly processor which would be able to deal with situations like this?
Thanks

[Brian Bushnell]

70bp overlap with 100% identity should not be joined, in general. That's too short and would cause rampant misassemblies if you instituted it as a general policy. Besides, do you know that the reference in this case is actually correct?

I could be wrong, but I thought that recent versions of PBJelly did try to join contigs that had sufficient support.

I wrote a program called Dedupe that will find all of these overlaps:

dedupe.sh in=contigs.fa am=t ac=f fo=t mo=70 ngn=f dot=graph.dot

This will find and print all the overlaps of at least 70bp with 100% identity (annotated by the overlap length, coordinates, and number of substitutions/edits). You can allow a fixed number of edits or substitutions in the overlap region with the "e=" or "s=" flag.

It won't merge anything, though it would be nice if someone wrote a post-processing program that used the overlap information to do merging. Still, 70bp is too short.