Hello,

Our objective is to generate a de novo assembly of the samples of our population. To do this we want to used ONT Simplex data, which was generated with a different objective (SV detection), using the library prep. guidelines suited for SV detection:

• Elimination of short DNA fragments using SFE kit

• Fragmentation of DNA using G-Tubes

This leads to us to the following R10 data:

• 121 Gb

• N50 = 13 Kb

• 47X coverage (genome size 2.6 Gb)

Of course, due to the use of SFE+G-Tubes, we lack longer read outliers. I understand not having these might complicate _de novo_ assembly, however we thought that having 99% coverage of the reference genome and a good depth would overcome this limitation.

Anyway, this is the pipeline that I have used for the _de novo_ assembly:

1. Base-calling using using sup model

2. Elimination reads with a length shorter than 5Kb and Q less than 15

3. hifiasm to generate the contig-level aseembly

When I look at the QC of the contig-level assembly I see that we have short contigs:

• N50: 250 Kb

• Completeness 99% (but 55% of duplicated genes)

1. Long-read polishing

2. Short-read polishing

3. Reference-based scaffolding

When I do the reference-based scaffolding is where I have problems. While the reference chromosomes are close to 100% covered, our de novo chromosomes are too large. To the point that the largest chromosome is 30% longer than reference. Of course this is biologically false. It looks like the short contigs lead to overlaps that cannot be resolved, leading to a slow and steady elongation of the chromosome. See the attached pictures:

I was wondering if there is any chance to modify the parameters of hifiasm to improve this situation, or if anyone here might know any additional step that might fix this issue.

Sorry for the long message,

Ferran.