telomeres getting lost

[JWDebler]

Describe the bug
I was running a comparison of current long read assemblers and realised that NextDenovo has one of the best contiguities, however, it seems to trim a lot of the telomeres.

Genome characteristics
40 mb fungal genome, haploid, about 20% repeats

Input data
Nanopore Q20+, 90x raw read coverage, read N50 ~15 kb

Config file

job_type = local
job_prefix = 15CUR005.nextdenovo
task = all
rewrite = yes
deltmp = yes
parallel_jobs = 4
input_type = raw
read_type = ont # clr, ont, hifi
input_fofn = 15CUR005.fofn
workdir = 15CUR005.nextdenovo

[correct_option]
read_cutoff = 1k
genome_size = 42m # estimated genome size
sort_options = -m 50g -t 30
minimap2_options_raw = -t 4
pa_correction = 5
correction_options = -p 30

[assemble_option]
minimap2_options_cns = -t 8
nextgraph_options = -a 1

NextDenovo
2.5.2

Flye vs NextDenovo
This is what the 5' ends of the contigs look like assembled by Flye 2.9.2 vs NextDenovo 2.5.2

Are there any parameters that can be tuned to avoid the loss and following manual curation of chromosome ends?
Cheers,
Johannes

[moold]

It's hard to say, in general, the default parameters set is best in most cases, I think you can try to merge the two assemblies first. Otherwise, you need to first confirm whether the corrected reads contain telomere reads, and then adjust the nextgraph parameters.

[JWDebler]

Thanks.
Where can I find the NextDenovo corrected reads?
I couldn't spot them in the output directory.

[moold]

Should be in 01_rundir/02.cns_align/01.seed_cns.sh.work/seed_cns*/cns.fasta

[JWDebler]

Thanks, found them. And yes, telomeres are missing from those corrected reads.

[webbchen]

I've observed the same issue in two oomycetes, most telomeres went AWOL. @JWDebler : Did you solve it eventually or did you just use another assembler?

[JWDebler]

@webbchen I did not. My workaround is that I assemble with both Flye and NextDenovo. ND seems better at assembling big repeat rich regions which Flye breaks into separate contigs. And Flye tends to be better around chromosome ends. I then use the ND assembly to manually scaffold the Flye contigs.