Hi,

I do have two drosophila samples, sequenced with the ONT technology. The aim of that sequencing is to compare these two samples, specifically on the chromosome X where I focus on a specific transposable element. Here are the metrics of the ONT reads :

file                                      format  type   num_seqs         sum_len  min_len  avg_len    max_len     N50
EBONT020_IsoR1_Femelles.fastq.gz          FASTQ   DNA   1,455,305   9,861,418,122        1  6,776.2  1,647,165  20,325
EBONT027_IsoR1deltaenv_Femelles.fasta.gz  FASTA   DNA   3,416,405  17,357,948,617        1  5,080.8    538,225  14,316

It is 75x and 133x of coverage, which is pretty unbalanced.

In order to identify a potential strucural variation in the region, I did :

1- Assembly with flye + racon (2 rounds) of both samples + dotplot :

file                                                  format  type  num_seqs      sum_len  min_len    avg_len     max_len         N50
EBONT020_IsoR1_Femelles.racon_iter2.fasta.gz          FASTA   DNA        370  138,540,524      273  374,433.8  21,434,776  19,757,405
EBONT027_IsoR1deltaenv_Femelles.racon_iter2.fasta.gz  FASTA   DNA        390  139,247,627      390  357,045.2  24,642,205  15,343,065

I did a dotplot of both assemblies. I specifically look at the chromosome X, where a potential long tandem duplication (>300kb) could happened in EBONT027_IsoR1deltaenv because of a CRISPR manip. In red the contigs corresponding to the chromosome X. But no duplication at first sight.

2- Mapping of the reads on the dm6 reference genome with minimap2

The sample 2 (EBONT027_IsoR1deltaenv) has a x2 coverage compared to the sample 1 at the positions between the two black lines.

Here is the duplication structure I suppose that happened :

I did a small python script to confirm that x2 coverage, taking in account the samples coverage (75x and 133x). I expect a coverage ratio of 1.77 (133/75) all along the genome, and ~3.4 when a duplication could happen. I highlight that with a plot, and I put in red the ratio when it gets higher than 3.4 :

It matches what I see in IGV. But I would prefer go with an official SV caller to highlight it.

3-SV calling

I did use 4 long-reads SV callers (svim, sniffles2, cutesv and spectre). I merged all the results but nothing came out.. What's interesting is that spectre gives coverage profiles : we can spot the x2 coverage in the sample2 (in the red box at ~0.16->0.2 = 1.6->2Mb). But the tool does not highlight the duplication in the output vcf file.

Is there something I do wrong? Sorry for that long post, but I am stucked. I also opened an issue on the spectre github (https://github.com/fritzsedlazeck/Spectre/issues/63)

Best