L2 estimate instead of LD matrix (R) for fine mapping summary stats

[kvshams]

With the availability of large LD matrix in LDSC format from UKBiobank (UKBB), is there any scope for susie to add function for using L2 estimate instead of R for the summary stats fine map as using actual LD matrix from UKBB is not practical for most of the users?

[gaow]

@kvshams not sure what you mean by L2 estimate -- I take that you are referring to the format LDSC stores LD matrices, and not the L2 norm used in LD score regression? Unlike LDSC, a whole genome regression of polygenic model with a focus on heritability estimation, fine-mapping typically involves much smaller candidate regions. LD matrix of UKBB scale of sample can be computed on regions of eg 3000 SNPs using tools such as PLINK or LDstore (not using R directly), and saved as plain text for susieR to use. So we don't have in mind a specific "standard" format to adopt for LD matrices for susieR. Of course I assume you can always add some codes in your pipeline to convert any LD matrix format (including what LDSC does) to a small, compressed text file, to load in R and analyze with susieR.