Hi al, I am trying to identify highly differentiated SNPs via window approach! I have 2 questions: first shall I try sliding window or adjacent windows? I wonder if there is any reference to help me to figure out which window size should I try?
I would appreciate if you know any appropriate reference to point me in that direction! Thanks
I believe that you are referring to linkage disequilibrium (LD) here.
People generally go for a sliding window of 50 SNPs per window and shifting the window by 5 SNPs each time, and then calculate (per window) the variance inflation factor (VIF) and LD for each batch of SNPs. LD is measured in different ways - take a look here at this Biostars thread: linkage disequilibrium: difference between D' and r-squared
PLINK can be used to measure this for you, assuming your data is in PLINK format: https://www.cog-genomics.org/plink/1.9/ld
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
I'm curious why do you favour a sliding window approach, this should lead to a lot of overlapping results wouldn't adjacent windows be more efficient?
Yes, it generates more information than the adjacent window approach, however, the adjacent window approach has a major flaw in that it only looks at blocks and therefore won't give me information on linkage disequilibrium (LD) between the blocks.
What if I have SNPs ranging from 1-20 and I analyse these in adjacent block sizes of 5 [SNPs]. I will not get information on LD between SNPs 1-5 and those in each other block. If I use a sliding window size of 5 SNPs and shift the window by 1 SNP each time, however, I will get a more continuous feel of LD across these 20 SNPs.
The ideal situation is to actually use a variable-sized sliding window based on SNP density in relation to genomic distance.