Replicating Hail's GWAS Tutorial

I decided to work on https://github.com/pystatgen/sgkit/issues/88 by trying to implement the Hail GWAS Tutorial with sgkit.

I’ll update this issue with my experiences. I barely know what I’m doing, so this should be fun.

[tomwhite]

With #471, it’s possible to create a histogram of DP values:

vcf_to_zarr(VCF_FILE, "1kg.zarr", format_fields=["DP"])
ds = sg.load_dataset("1kg.zarr")
dp = ds.call_DP.where(ds.call_DP >= 0) # filter out missing
xr.plot.hist(dp, range=(0,30), bins=30)

[tomwhite]

• The datasets have a substantial number of variables, which causes Xarray to collapse them in the HTML view, which means that new users might miss them. It would be nice if we could control this better. Also, the dataset attributes (that we are not so interested in) are shown expanded - it would be better if they were collapsed.

Now that https://github.com/pydata/xarray/pull/5126 is in, when the next version of Xarray is released (0.17.1), we can put the following at the top of the notebook to get the effect we want:

xr.set_options(display_expand_attrs=False, display_expand_data_vars=True)