trouble shooting PB assembly generating a larger than expected contig

[nwfungi]

So I've got a newly assembled genome that was sequenced using PacBio sequencing and obtained >100x coverage. It was assembled with HGAP.3 and for the most part looks great except for an exceptionally large contig (the largest actually). it's estimated that our genomes largest chromosome is around 3.5Mb (based on electrokaryograph) but this one large contig is around 5Mb and has low complexity across the entire contig. Blasting this contig doesn't result in any matches of note at NCBI or to our reference genome.

Any suggestions on how I could adjust the assembly to get rid of this large contig that I am fairly sure is not real?

[Brian Bushnell]

Why don't you try gene-prediction on it to see what you get? Also, based on mapping, what kind of coverage does this contig have? Note, also, that it could be a bacterial symbiont, so you might want to try prokaryotic gene-calling. Bacteria don't usually have low complexity, though.

[rhall]

Looking at the coverage when remapping all the raw data would be the most telling, does it have 100x coverage, is the coverage even?
I'm actually really intrigued, I've done a lot of HGAP.3 assemblies, but have never seen a 'junk' contig get anywhere near that big. It's possible its just all the low complexity repeats getting overlapped together, but normally this would generate at max 10's of kb of sequence. You can also look at the overlap graph to see what the origin of the contig is https://gist.github.com/rhallPB/2d962e700d83270b0109 .

[gringer]

Do you have Illumina reads that you can map to this contig in local mode (e.g. 'MagicBLAST' or 'Bowtie2 --local')? If not, you could try digitally fragmenting your PacBio reads into short reads and mapping.

Map only to the single contig. As rhall has said, you should get a somewhat even coverage across this contig. Any big jumps in coverage indicate something that needs further investigation. A shift from one coverage level to another might indicate a misassembly, while a blip of extremely high coverage suggests transposon sequence that may be interfering with assembly.