How to generate ld plot for 100000 SNPs in LD?

Question

[mandecent.gupta]

Hi,

I have long list of SNPs which are in LD to each other (almost 100000) when I try to generate LD plot using heat map. It takes lot of time for generating the plot. May I know is there any utility which optimize this plotting for 100000 records.

Input list (millions of records)

Coord_X      Coord_Y      intensity
1                      2                 0 .1
2                      9                 0.4
3                      11               0.5
4                      13               0.5
5                      15               0.1
..
..
..
..
10000000      10000000   0.8

Regards

Answer 1

[Philipp Bayer]

I wouldn't make a plot this large because it won't tell you much, you'll just have a gigantic picture that'll take you weeks to fully comprehend.

Some things you can however try:

1. Merge adjacent SNPs with LD=1 (or >0.9 etc.) into tagging SNPs (TagSNP, and if I remember correctly Haploview can do that). You'll also get the regions with very high LD this way.
2. Plot LD only for windows of 10KB (20KB...), not for all against all SNPs. That's why you usually see these long vertical plots with smaller triangles all over the place instead of one gigantic triangle for all against all SNPs
3. Plot it just for your regions of interest - any genes? Or you write a small script that extracts interesting regions of strange or interesting LD and then plot these regions only