Canu sensitivity to genomeSize parameter

[jpummil]

I am currently working with some Bio faculty on a de novo assembly of Centruroides vittatus (Striped Bark Scorpion). One difficulty I seem to be having involves the command line parameter "genomeSize=" in Canu. Of course, one of the real difficulties with de novo is you likely don't know the size of the genome. You may not even have an estimate from a close relative species.

I currently have two PacBio runs (different specimens in each) labeled Q1133 and Q1171. It seems that by varying the "genomeSize=" parameter, the output of the assembly can be significantly different! I am attaching a Quast output for some recent runs...two for each data set, with only the genomeSize parameter modified (genomeSize=170 and genomeSize=350).

It seems that not many people have produced tools to estimate genome size in new projects that involve PacBio or ONT data...I did have some old Illumina data that was kind of iffy, but I tried the new GenomeScope tool from Schatz, et al and it comes up pretty consistently around 180Mb for an expected genome size. It was suggested to try the tool using the corrected reads out of Canu, but the tool doesn't seem to be able to evaluate that sort of input at this time.

Any suggestions on how to best approach the problem (if it IS a problem and not simply an error on my part)? I am using essentially default values in Canu.

[rhall]

Now that you have some contigs, why not map all your raw data to one of the assemblies and look at the coverage distribution across the contigs.