Jittering and additional peaks

[Tami Hildahl]

Good morning.

I have a dataset of around 10,000 SNPs. When my file is read, it recommends jittering because several markers are in the same position. I have run through both with and without jittering for comparison to find that I receive extra peaks on an additional linkage group while jittering but only for standard interval mapping where the lods are far larger than the other analyses. Is jittering necessary for SNP data, and why would I only see the additional peaks with standard interval mapping when I have also used imputation and HK? The first plot is without jittering, the second is with jittering.

Please advise.

Tami

[Karl Broman]

I wouldn't expect jittering to have such a large effect. Do those markers at the same position have genotypes which are consistent with each other?

For technical reasons, we don't want to have inter-marker distances of 0, and that's what jittering fixes.

The extremely large peaks in the right plot for interval mapping look like the sort of false peaks that can occur when you have a lot of missing data and a phenotype with a bimodal distribution. (See Fig 4.14 and the related discussion in the R/qtl book, https://rqtl.org/book/rqtlbook_ch04.pdf) But I wouldn't expect that to occur when you have such dense markers.

karl

--
You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rqtl-disc+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/rqtl-disc/1eac9f3d-1579-4fbf-be0b-796a42220250n%40googlegroups.com.

[Tami Hildahl]

Thank you, Karl. 

I do have another question. The LOD significance thresholds are genome-wide. Is there a way to obtain chromosome-wide thresholds? I didn't see it noted in the manual. 

Cheers,

Tami

[Othmane Lamoumni]

Hello everyone,

For the chromsome-specific thresholds, isn't that achievable using the "chr" argument in scanone?

Perm_t <- scanone(cross, pheno.col = 1, chr = 1, model = "normal", method = "hk", n.perm = 2000)

Othmane

Le jeudi 14 mars 2024 à 10:52:19 UTC, Karl Broman a écrit :

No, R/qtl has no facility for calculating chromosome-specific thresholds.

karl

[Karl Broman]

Ah, very good. Yes, that would work.

thanks,

karl

To view this discussion on the web visit https://groups.google.com/d/msgid/rqtl-disc/94ce499d-a541-49f1-9b6f-88c405ce09bbn%40googlegroups.com.