Merge assembled contigs?

[donkirkby]

Sometimes, the IVA assembler can't assemble one large contig, but it assembles several smaller contigs that either have small gaps between them or have some overlap.

For example, 73060A-HCV_S46 from 15 Jul 2016 has a whole-genome contig for HCV-2b, and several HCV-1a contigs with overlaps of hundreds of bases.

When I looked at the IVA source code, I found that it requires 99% identity in the overlap region before it will merge two contigs. The 73060A sample had overlaps with between 96% and 99% identity. When I lowered the requirement to 95%, it combined the HCV-1a contigs nicely, but it still reported a separate contig for HCV-2b.

There are a few other samples that had trouble combining contigs, but I haven't looked at them in detail yet. HIV3428P100IN200-C19-HIV-S51 from 20 Sep 2019 run is the closest, but it looks like one contig has primer at the end. Samples HIV0887-P2D21-HIV_S3 and HIV0887-P2C12-HIV_S32 from 30 Aug 2019 looks even better, but have very little overlap.

For now, Chanson and I decided to leave it as is, and MiCall will combine the nuc and amino counts for overlapping contigs in later steps. In a future version, we should look to see how often we have trouble merging contigs, and how many of them have this overlap identity problem.

[CBeelen]

There are two places that the sequence identity requirements are set in IVA - once when run_nucmer() is called to find overlapping regions (where is it currently set to 95%), and once again when to_graph_edge() is called to add additional edges (of overlapping contigs) to the assembly graph. There is it currently still set to 99%, and the minimum overlap length is set to 200 (both by default). Changing these defaults or calling to_graph_edge() with different inputs might solve this issue. I'm currently experimenting with this in the Charlotte branch of cfe-lab/iva.