Number of CSs

[Xi-Cao]

Hi there,
I'm confused that how the CSs were selected. I thought the order min_abs_corr was used to determine the min corr of SNPs incorporated into a CS.


While when I tested in another locus, using this command didn't change the number of CSs it originally found. So why didn't it capture some new CSs with lots of SNPs, that have very small min corr?

Thanks，
xicao

[pcarbo]

@Xi-Cao min_abs_corr filters out CSs that do not satisfy the minimum absolute correlation ("purity") criterion; it has no effect on the ordering of the CSs.

If the result does not change, it means that all CSs satisfied the previous purity criterion.

[Xi-Cao]

Thanks for your reply.
So susieR found 9 CSs in the first example, while some of them didn't meet the purity criterion. In the next example, it actually found only one CS. Did I get this right?
I'm confused that why sometimes multiple CSs, each of which contains almost the entire snp with small purity of the region, are found in a region, while some regions just have few CSs. Do the PIP and absolute correlation determine the number of CS, and how?

[Xi-Cao]

By the way, considering it affects the output of CSs, is there a recommended min_abs_corr value or a selected method？

[pcarbo]

The basic idea is that there should be one CS for each independent signal (causal SNP), and different regions will have different numbers of causal SNPs.

Since susie sometimes outputs a CS containing a very large number of SNPs (each with small PIP), the "purity" filter is helpful for removing these CSs that are likely not interesting. Therefore, we recommend setting "min_abs_corr" to a number larger than zero, but we do not have a recommendation as to the exact number. (The default min_abs_corr = 0.5 often works well.)

[Xi-Cao]

I get it, thank you!😊