Differentiation index and GWAS/LD analysis for rs1233572 - revised 2026-10-31

DIFFERENTIATION INDEX (descriptive; 7 gnomAD ancestry groups): rs1233572 = 0.0771
  Definition: unweighted variance of per-population alternative-allele frequency divided by
  the heterozygosity of the pooled, sample-size-weighted frequency. This is NOT the Hudson or
  Weir-Cockerham estimator and is NOT a test for selection.
  per-population alt(C) freq: afr .872  eas .628  amr .589  asj .589  sas .587  nfe .489  fin .396

BACKGROUND COMPARISONS: ATTEMPTED, NOT REPORTED IN THE MANUSCRIPT
  The 2026-10-30 version of this file reported rs1233572 at the 36th percentile against 44
  matched variants in the surrounding extended MHC and the 76th percentile against 29 matched
  variants on chromosome 2, and concluded "no evidence of selection". That conclusion is
  withdrawn. Two faults were found.
  (1) Comparators were matched on ALTERNATIVE allele frequency over a 0.25-0.55 window, which
      excludes the target's own alternative frequency of 0.610. Matching must be on minor
      allele frequency (target MAF 0.390).
  (2) With MAF-matched comparators the percentile is unstable to the sampling window. Four
      windows, same method, same target:
        xMHC  chr6:28.3-29.2 Mb   n=22  background median 0.110  -> 18th percentile
        xMHC  chr6:27.6-30.4 Mb   n=22  background median 0.059  -> 95th percentile
        chr2  100.0-100.9 Mb      n=22  background median 0.029  -> 95th percentile
        chr2  98.0-101.6 Mb       n=55  background median 0.146  ->  9th percentile
      The background median varies about five-fold between sub-intervals, so the percentile
      reports the choice of window rather than a property of the variant.
  The manuscript therefore reports the index alone, descriptively, and draws no inference about
  selection. Testing for selection at this locus would require haplotype-based statistics on
  phased genotypes, which were not available to this analysis.
  Script: analysis_scripts/fst_maf_matched.py (MAF-matched version); the superseded
  alternative-frequency version is analysis_scripts/fst.py.

GWAS / LD:
  rs1233572 direct GWAS associations: 0
  LD partners at r2>=0.6 (CEU/YRI/CHB/FIN): 260
  GWAS associations involving that LD block: 39, across 28 traits
  GWAS associations within chr6:28.60-28.90 Mb: 527, across 241 traits (~5x genome-average density)
  Locus lies INSIDE the extended MHC (~chr6:25.7-33.4 Mb).
  CONCLUSION: causal attribution is not possible without fine-mapping. The GWAS route is blocked
  at this locus.
  Source: GWAS Catalog full release (1,203,840 associations), Ensembl REST LD. Queried 2026-10-30.
