Running LDSC with GRCh38/hg38 annotations

[Peter Hickey]

Hi,

I have annotations with GRCh38/hg38 co-ordinates for which I would like to 'partition the heritability' of GWAS traits using LDSC.

I have previously done this using annotations mapped to hg19 (https://www.ncbi.nlm.nih.gov/pubmed/30643296).

Based on https://data.broadinstitute.org/alkesgroup/LDSCORE/readme_baseline_versions, I should now be using the v2.2 baseline model for this task.

Initially, I used liftOver (via the Bioconductor 'rtracklayer' package) to convert all my annotations from hg38 to hg19 and the re-ran the same pipeline I used in the previous paper.

However, I'd love to do everything using hg38 co-ordinates  and I recently noticed that some GRCh38/hg38 LDSC files have been made available (https://data.broadinstitute.org/alkesgroup/LDSCORE/GRCh38/).

I've started re-writing my code to migrate to the hg38 version of LDSC (as well as move to v2.2) but have a places where I'd appreciate some guidance.

Firstly, https://data.broadinstitute.org/alkesgroup/LDSCORE/GRCh38/ doesn't seem to contain all the files I need for my analysis.

Here are some files used in my previous pipeline that I have been unable to find hg38-equivalent/compatible versions of:

- https://data.broadinstitute.org/alkesgroup/LDSCORE/1000G_Phase1_cell_type_groups.tgz

- https://data.broadinstitute.org/alkesgroup/LDSCORE/w_hm3.snplist.bz2

I've also only found the `baselineLD.*.l2.M` and `baselineLD.*.l2.M_5_50` files in https://data.broadinstitute.org/alkesgroup/LDSCORE/baselineLD_v2.2_bedfiles.tgz, which doesn't seem to be hg38-specific.

Are there hg38 versions of these files available and/or are they needed?

Secondly, Is it necessary to liftOver all the GWAS summary stats to hg38?

I read a few posts on this forum and it seems like the effect of using hg19 GWAS summary stats is minor (https://groups.google.com/forum/#!searchin/ldsc_users/hg38|sort:date/ldsc_users/_wIQrqK57Nc/ur6l2GH0AgAJ), although I do worry about a reviewer fussing over this.

Thanks for your help,

Pete

[Peter Hickey]

Bump.
Is the Google Group still active or is there somewhere/someone else I might ask?

Thanks,

Pete