Pooled samples and mosaic option

[SheepwormJM]

Hi,

Just following on from the 2017 question 'Detection on Pooled Samples #21 '.

Would using the --mosaic option be useful for pooled sequencing data? Or would you recommend running the standard calling method?

Thanks,
Jenni

[SheepwormJM]

Update to above: I ran the mosaic and found that it missed some very obvious SVs in the data (visualised in bam files on JBrowse2), and seemed to be mis-identifying others. Running it without mosaic identified fewer SVs in one sample and a lot more in another, and seemed to better identify a deletion that I had thought was incorrectly identified before.

So, I am thinking that the mosaic option is specifically looking for low frequency SVs, but I'm not sure why it would incorrectly identify them.

I will keep on looking at the data and look to see what else I can tweak. If anyone has any insights would be pleased to hear your advice! Thanks.

[fritzsedlazeck]

Hey @SheepwormJM
Sniffles 2.7 coming out today...give it a try.. we improved the mosaic filtering as there were some bugs in there. One the mosaic SV benchmark from SMAHT this improved the performance significantly.

The mosaic option is indeed only looking at low VAF variants (<22% VAF) . So the combined mode that is now fixed in 2.7 would provide all VAF ranges (including germline).

Thanks
Fritz

[SheepwormJM]

Thanks @fritzsedlazeck will give it a go now.

[SheepwormJM]

Hi,

I ran the new Sniffles2 v2.7.0 with the following command on my poolseq ONT data:

sniffles -i $SAMPLE2_BAM --snf ${SAMPLE2_NAME}.snf -v ${SAMPLE2_NAME}.vcf --mosaic-include-germline --tandem-repeats $TANDEM_REP --reference $REF_FA --output-rnames --all-contigs

which ran fine and provided a similar total of SVs for each of my samples as when I had run the previous version with the mosaic and standard options separately.

However, looking at the output I am now missing a very large (and complex) and common SV from my two samples. Looking at aligned bams to the genome I can see it is present in both, but possibly at too low a frequency to be picked up in one samples, and it wasn't picked up in this sample last time either.

But it had 153 reads supporting it last time in the other sample and is not present this time, but another less common one is.

# Sniffles v2.6.3
160325  Sniffles2.DUP.596S6     G       <DUP>   56      PASS    PRECISE;SVTYPE=DUP;SVLEN=4021;END=164346;SUPPORT=153;RNAMES=

166430  Sniffles2.INS.24S6      A       ATGATTGACGGCTACATTTCGATTAATTTCTGATGTTTGTTCTCTTTGTCTCCTGATGAT    60      PASS    PRECISE;SVTYPE=INS;SVLEN=59;END=166430;SUPPORT=31;RNAMES=

# Sniffles v2.7.0
# No variant for 160325 for either sample
166430  Sniffles2.INS.16S6      A       ATGATTGACGGCTACATTTCGATTAATTTCTGATGTTTGTTCTCTTTGTCTCCTGATGAT    60      PASS    PRECISE;SVTYPE=INS;SVLEN=59;END=166430;SUPPORT=31;RNAMES=

Any ideas?

Thanks,
Jenni

[SheepwormJM]

Just ran it without the mosaic combination option and the common SV above is still missing. It is a complex SV, but it is also massively increased in read coverage relative to most of the rest of the genome.

I'm also missing another duplication which had a read support of over 1000 as detected by v2.6.3 too.

In context, the genome average depth per site is only 44.

[fritzsedlazeck]

Thanks Jenni,
yeah we indeed readjusted filters in the new version. Can you share a part of the bam file with us? We dont have pooled sequencing data .. also you could try to use --no-qc to obtain all candidate SV. This is a debug option for us... but some people like to use it too..

cheers
Fritz

[SheepwormJM]

Thanks for sticking with this Fritz!

Just tried the --no-qc and got back almost all SVs the log file says it put in the snf. I can now see the variants I've mentioned. One has a huge number of reads supporting it, and is called 'PRECISE' but also has a several other SVs at the same base. While the other, more complex one (I think, definitely complex anyway, but haven't actually looked at the bam for the other, only read coverage ratio), has been split into two based on SVLENGTH, one IMPRECISE and the other PRECISE.

How is best to share the bam file with you? Do you need the reference etc too?

Jenni

[SheepwormJM]

I think a lot of the ONT reads are quite short too, not sure if that helps pick things apart.

[fritzsedlazeck]

yeah .. the filter tag in the VCF file can show you why (if) Sniffles rejected them..

I assume its not human ? then yes reference file please. and if you cut out the region of the bam file you probably can just email it over : fritz.sedlazeck@gmail.com
Thanks
Fritz

[SheepwormJM]

Yes, not human! :) Thanks. I'll get it extracted and sent to you.

Jenni

[lfpaulin]

Thanks for the additional information @SheepwormJM I out latest development version and found the DUP you mentioned marked with a filter and tagged as mosaic, which does not make much sense

qqPsoOvis1.1_s07	160325	Sniffles2.DUP.31S6	N	<DUP>	56	ALN_NM	PRECISE;MOSAIC;SVTYPE=DUP;SVLEN=4021;END=164346;SUPPORT=129;RNAMES=...;COVERAGE=56,1078,1243,886,47;STRAND=+-;NM=0.088;STDEV_LEN=0.000;STDEV_POS=0.000;VAF=0.120	GT:GQ:DR:DV	0/0:60:950:129

we had an update in the coverage computation in v2.7, so it is likely there are some issues here as we have not checked so large coverage differences (we don't have such an example for testing)

We will keep you posted

[SheepwormJM]

Thanks Luis!

…

_________________________________ Please note that I don’t expect you to read or respond to this email outside your normal working hours, or when on holiday. I will not normally check emails outside of working hours. I currently work Monday to Thursday and will aim to respond to you as soon as I can when I am in the office. Dr Jennifer McIntyre Research Associate School of Biodiversity, One Health and Veterinary Medicine College of Medical, Veterinary and Life Sciences University of Glasgow Charity number SC004401 Rm 242, Henry Wellcome Building University of Glasgow Garscube Estate G61 1QH Work telephone: +44(0)141 330 8216 / 07791 644 116 Github: https://github.com/SheepwormJM<https://github.com/SheepwormJM?tab=repositories>

________________________________ From: Luis Paulin ***@***.***> Sent: 24 November 2025 22:48 To: fritzsedlazeck/Sniffles ***@***.***> Cc: Jennifer McIntyre ***@***.***>; Mention ***@***.***> Subject: Re: [fritzsedlazeck/Sniffles] Pooled samples and mosaic option (Issue #590) [https://avatars.githubusercontent.com/u/17764146?s=20&v=4]lfpaulin left a comment (fritzsedlazeck/Sniffles#590)<#590 (comment)> Thanks for the additional information @SheepwormJM<https://github.com/SheepwormJM> I out latest development version and found the DUP you mentioned marked with a filter and tagged as mosaic, which does not make much sense qqPsoOvis1.1_s07 160325 Sniffles2.DUP.31S6 N <DUP> 56 ALN_NM PRECISE;MOSAIC;SVTYPE=DUP;SVLEN=4021;END=164346;SUPPORT=129;RNAMES=...;COVERAGE=56,1078,1243,886,47;STRAND=+-;NM=0.088;STDEV_LEN=0.000;STDEV_POS=0.000;VAF=0.120 GT:GQ:DR:DV 0/0:60:950:129 we had an update in the coverage computation in v2.7, so it is likely there are some issues here as we have not checked so large coverage differences (we don't have such an example for testing) We will keep you posted — Reply to this email directly, view it on GitHub<#590 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ANH2WOT5RFRQ5EXYQAFAT6336ODKLAVCNFSM6AAAAACIRGFJT2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKNZTGA2DAMZZHA>. You are receiving this because you were mentioned.Message ID: ***@***.***>