sample and phenotype preprocessing for aou v8

[aryarm]

some differences between v7 sample preprocessing and v8 sample preprocessing:

1. v7 used the number of singletons while v8 used a count of the number of variants not appearing in gnomAD 3.1 instead
2. v7 used the Het/HomVar genotype ratio for all genotypes while v8 has separate het/homvar ratios for SNPs and indels
3. v7 filters samples for call rate > 0.9 while v8 does not perform any filtering for call rate. We should filter for call rate before each GWAS

sample QC report

• started out with 414831 samples
• our filtering step left 380247 samples (removed 34,584 samples)
  • of those 34,584 samples, AoU had flagged 988 of them
• v8 passing samples are missing 9436 samples from v7 that passed
  • and originally, the v8 data (before even any qc) was missing 2684 samples from v7
• overlapping them:
  • how many samples in v7 before qc? 245395
  • and how many in v8 that overlap v7? 242711
  • that is a loss of 2684 which matches from before!
• after running the v8 code with only the v7 samples from the v8 metrics (aka v8code+v7subset). what does the passing data look like?
  • comparing v8code+v7subset with v7passing:
    • how many unique to v8code+v7subset? 4910 samples
    • how many unique to v7passing? 9442 samples
• ok, now let's use the v7 metrics to run finish_sampleqc.py with the v8 code (aka v7metrics+v8code)
  • comparing v7metrics+v8code with v8code+v7subset:
    • how many unique to v7metrics+v8code? 7 samples
    • how many unique to v8code+v7subset? 87 samples
  • comparing v7metrics+v8code with v7passing:
    • how many unique to v7metrics+v8code? 4823 samples
    • how many unique to v7passing? 9435 samples