Genome Size Estimation from PacBio Raw Reads

[jpummil]

So, working on a de novo assembly using Canu, and it seems to be VERY sensitive to the genomeSize=XXX parameter which is required. As it is a new project, no one has an actual "size" on it (checked T Ryan Gregory's site...nothing similar there either).

So, I am using BBMap suite, specifically...the "kmercountexact.sh" component. Waiting on a compute node right now with >64GB of ram to run, but have it set as follows: kmercountexact.sh in=filtered_subreads.fastq khist=khist.txt peaks=peaks.txt out=genomesize.txt

As Brian Bushnell is active on here, I was hoping to inquire about using this on PacBio specifically...anything I need to be more specific about on the options? Also, can I specify both of my PacBio files as arguments? I have both a .fastq of the long reads as well as a .fasta of much shorter reads supplied by the sequencer people. I know it can do PE files as in= and in2=, but what about to essentially "single" reads?

[Brian Bushnell]

Hi Jeff,

Unfortunately, I don't have a good method for this. I've tried kmercountexact, and it does not work on raw PacBio reads due to the high error rate. I do not know of a better method for genome size estimation than assembling, with Falcon, for example. Sorry!

If you have multiple files, though, you can enter them comma-delimited, like this:

kmercountexact.sh in=filtered1.fq,filtered2.fq

Not all tools support that, but Tadpole, KmerCountExact, and Dedupe do.

[jpummil]

Thanks for the quick response, Brian!

Good to know about the comma-delimited method for multiple entries. Unfortunate to hear about the PacBio error issue when trying to determine genome size. I thought about this a bit and am wondering if the pre-processing Canu does to the data could be used prior to trying kmercountexact? It outputs a couple of files during its run which trim, then correct the reads:

<filename>.trimmedReads.fasta.gz

then

<filename>.correctedReads.fasta.gz

Of course, they have been processed WITH the genomeSize estimate provided at run time and I'm not certain of how much that might have influenced any trimming or correction. I might try and contact Phillippy or Koren and inquire further ;-)

[wdecoster]

Perhaps a quick and dirty assembly with miniasm can give you an idea? https://github.com/lh3/miniasm

[jpummil]

Perhaps a quick and dirty assembly with miniasm can give you an idea? https://github.com/lh3/miniasm

Thanks for the suggestion wdecoster I think I've avoided miniasm thus far because it appears to only output .gfa files? Kind of limits further evaluation of the assembly as most common tools seem to still only take .fasta.

Update: Found a note in another thread about converting .gfa to .fasta Trying it now...

awk '/^S/{print ">"$2"\n"$3}' in.gfa | fold > out.fa

[jsoghigian]

Thanks for the suggestion wdecoster I think I've avoided miniasm thus far because it appears to only output .gfa files? Kind of limits further evaluation of the assembly as most common tools seem to still only take .fasta.

Update: Found a note in another thread about converting .gfa to .fasta Trying it now...

awk '/^S/{print ">"$2"\n"$3}' in.gfa | fold > out.fa

About to try this method myself - jpummil, were you successful in estimating genome size from your raw reads?

[jpummil]

About to try this method myself - jpummil, were you successful in estimating genome size from your raw reads?

The assembly itself using miniasm and the conversion script from gfa to fasta worked fine, though the assembly isn't as "good" as from Canu.

Still no really good way to estimate a genome size from the PacBio reads. Schatz put together a really nice tool called GenomeScope, but currently only works with Illumina reads.

[Markiyan]

Try analysing the reads from the short inserts (multipass ones).

You can try extracting the long raw reads from the short library inserts, which pass the insert multiple times (CCS-like reads), doing self error correction, and than using kmer counter software designed for Illumina, 454 or Sanger data.

Also please be aware, that you may have to screen out the high copy number DNA (mitochondrial/plastid genomes) before doing kmer counting.

Also you may get some PCR-Free miseq data to complement your pacbio assembly. (Can be cheaper if your coverage is still too low).

[kartika]

thanks you