Open Huifang-Xu opened 1 year ago
@Huifang-Xu What is the purity of the CSs?
This is the purity of the small region (Nsnps = 800).
There is no output for the entire locus (Nsnps = 2953).
@Huifang-Xu Try setting min_abs_corr
to a smaller number, e.g., 0.
I reran the model for the entire locus, the min.abs.corr
of L1 is 0.295.
susie_rss_z_Purity <- susie_rss(sumstat$Z, R=ld_matrix, n=sampleSize, L = 10,min_abs_corr=0)
print(susie_rss_z_Purity$sets)
For the small region,
@pcarbo I wonder why the credible set of the small region has higher purity while the credible set of the large region has lower purity. Should we choose a small region for fine-mapping? e.g., 1000 SNPs or a 250Kb window.
@pcarbo I realized a distant SNP was included in the cs for the entire locus but not in the small one. I think that's why the large region has a low purity and does not produce cs. But again, should we choose a small region for fine mapping, e.g., 1000 SNPs or a 250Kb window?
Here are the SNPs included in the cs for the large region (Nsnp=2953):
Here are the SNPs included in the cs for the small region (Nsnp=800):
Thanks, Huifang
@Huifang-Xu In general, it is safer to choose a larger fine-mapping region. The results for this particular locus do not look particularly strong; this may be a "borderline" result.
@pcarbo I see. Thank you very much for your prompt reply and help! I really appreciate it.
Huifang
Should I always use susie_get_cs()? Because I usually use summary(susie_rss_z)$cs to extract and plot cs.
For susie_get_cs
, it requires X or Xcorr to output the same CSs as summary(susie_rss_z)
. It uses pairwise correlations to filter CSs. In the case X and Xcorr are unspecified, it outputs L CSs (L = 10 by default).
@zouyuxin I see. Thanks for the clarification.
Hi there,
I've been using
susie_rss()
for fine-mapping summary statistics using LD references from 1000GP. For one locus (Nsnps = 2953), I can't get a credible set fromsummary(susie_rss_z)$cs
, but I have been able to get a credible set fromsusie_get_cs(susie_rss_z)
. The PIP plot does not show cs, but the highest PIP is around 0.8.Here is my code,
susie_plot(susie_rss_z,y="PIP")
However, if I narrow the region to 800 SNPS, I can get a credible set from
summary(susie_rss_z)$cs
andsusie_get_cs(susie_rss_z)
. The highest PIP of the same SNP is similar.susie_plot(susie_rss_z_sub,y="PIP")
I want to know why there are different outputs. Should I always use
susie_get_cs()
? Because I usually usesummary(susie_rss_z)$cs
to extract and plot cs.Here are the summary statistics and LD matrix. I appreciate your help. https://github.com/Huifang-Xu/test_SuSiE/blob/main/locus1.zip
Best, Huifang