Test usefulness of adding kinship; Can one run permutation on independent chromosomes
14 views
Skip to first unread message
Jonathan Berlingeri
Mar 20, 2024, 6:00:56 PMMar 20
to R/qtl2 discussion
Hello Dr. Broman,
Thank you for this resource and your clear documentation. It has made my life easier. I am preforming mapping in an 8-parent MAGIC population with selfing (riself8). I have evaluated population structure with PCA (outside of qtl2) and have not found much structure at all, as expected in a population of this design. I would however like to verify the usefulness of adding kinship by looking at model fit with and without itas well as LMM vs. LOCO. I have compared scan plots made without inclusion of kinship and with it and the shifting in LOD is only nominal.
1) Is there a way to assess the fit of the models in qtl2? Or otherwise compare model performance?
Further, the chromosomes in my organism vary in substantially in size, gap distance (cM), and LD half decay. I want to ascertain whether or not there is something problematic occurring within any linkage group because of this variation (over 4-fold for half decay and ~2-fold for size).
2) Is there a way to permute on a independent chromosome basis?
I also noticed that one of my chromosomes has larger gaps in genetic distance between markers up to 6 cM, (most gaps are about 1 cM). I am hesitant to address this with pseudomarkers because if I discover a QTL that includes a pseudomarker I am not sure how that affects biological interpretation.
3) I found a reference to pseudomarkers in chapter 4 of your and S´aunak's guide to R/qtl. Is there any further description available?
Many thanks,
Jonny
Thank you for this resource and your clear documentation. It has made my life easier. I am preforming mapping in an 8-parent MAGIC population with selfing (riself8). I have evaluated population structure with PCA (outside of qtl2) and have not found much structure at all, as expected in a population of this design. I would however like to verify the usefulness of adding kinship by looking at model fit with and without itas well as LMM vs. LOCO. I have compared scan plots made without inclusion of kinship and with it and the shifting in LOD is only nominal.
1) Is there a way to assess the fit of the models in qtl2? Or otherwise compare model performance?
Further, the chromosomes in my organism vary in substantially in size, gap distance (cM), and LD half decay. I want to ascertain whether or not there is something problematic occurring within any linkage group because of this variation (over 4-fold for half decay and ~2-fold for size).
2) Is there a way to permute on a independent chromosome basis?
I also noticed that one of my chromosomes has larger gaps in genetic distance between markers up to 6 cM, (most gaps are about 1 cM). I am hesitant to address this with pseudomarkers because if I discover a QTL that includes a pseudomarker I am not sure how that affects biological interpretation.
3) I found a reference to pseudomarkers in chapter 4 of your and S´aunak's guide to R/qtl. Is there any further description available?
Many thanks,
Jonny
Karl Broman
Mar 21, 2024, 9:47:08 AMMar 21
to R/qtl2 discussion
1. I don't have a way to assess model fit. If the differences are nominal, then in my mind there's no concern.
2. To perform permutations separately for each chromosome, you could provide scan1perm() with data for a single chromosome.
operm <- vector("list", length(probs)
names(operm) <- names(probs)
for(chr in names(probs)) operm[[chr]] <- scan1perm(probs[, chr], pheno, kinship[chr], n_perm=1000)
3. By "pseudomarker" we just mean a position between markers that is considered as a putative QTL location. I don't know what sort of description you'd find useful; the idea of goes back to interval mapping introduced by Lander and Botstein (1989)
karl
Reply all
Reply to author
Forward
0 new messages