I'm trying to assemble a nematode genome that we estimate to be 400Mb in length. I have 60x worth of PacBio Sequel data and I've been using HGAP4 (FALCON + Arrow) via the SMRT Link gui. So far I'm not having much luck getting a good assembly and I'm trying to decide if its worth tweaking parameters more than I already have, or if it would be worth generating more PB coverage.
My primary metric for how good the assembly is has been BUSCO to look at core gene coverage and then simple contig metrics (N50, total length, # contigs, etc...) So far my best assembly has been missing 35 BUSCO core eukaryotic genes, and a fair number of the genes that were found were found > 1 time. Also the total length of contigs we're getting is > 100Mb larger than we expected (although its within the world of possibility that our genome size estimate is wrong). So I suspect this worm's genome may be highly polymorphic. We do have an old assembly but its in pretty bad shape itself (and the reason we want to build a new, better one).
Can anyone suggest arguments for HGAP4 that might help improve contiguity in a genome that is very polymorphic? So far I've tried HGAP4's default params, default params + the 'aggressive' switch, and a range of values for cutoff_length_pr. If I set cutoff_length_pr to just a bit < the mean pre-assembly read length my contiguity goes up and the genome size drops to closer to what I had expected, but I end up losing a few core genes. So I think its really just dropping data that is needed.
Any advice would be appreciated.
I'm no expert with HGAP4, but why not try an alternative assembler such as Canu. At least you'll have something to compare to. Also, if you have a diploid organism you might get some haplotype information out of the advanced Canu functions similar to Falcon's Unzip principle. This might resolve your genome size estimate issues.
This shouldn't be too tricky on a 400Mb genome.
Hi, re: BUSCO. I am working on plant genomes and 35 missing seems a fairly good result in terms of percentage. What is you total Complete BUSCOs in %? Maybe the assembly you get is not that bad? I am also curious about the time that it took to run the assembly with Falcon. I am trying to assemble now a pomegranate genome with Falcon. The genome is of similar size (~400Mb) and I am running it on 20-core machine with 504G memory. I have relatively high coverage (~120x) and it runs forever already. I am curious how long did it take to run in your case. THANKS
How long is "forever ?"
It runs more than one week. Around 7 days and 10 h.