Project 2 v0.14 - AHSG 3-Mb summary-statistic sensitivity

Source extraction: 14,507 rows retained from 13,615,309 source rows; 0 malformed rows.
Standardized complete rows: 14,507; biallelic SNVs: 13,532.
Window: chr3:185,119,924-188,119,924 (GRCh38; +/-1.5 Mb around rs4917).
Regional marginal lead: rs4917, P=3.35e-12.
Top five marginal variants: rs4917, rs4918, rs1900618, rs7635884, rs13073106.
Genome-wide significant SNVs: 11, all spanning 9,879 bp (186614393-186624271).
Best signal >50 kb from rs4917: rs775702472 at 187647186, P=2.35e-05.

Wakefield single-effect ABF sensitivity:
  W=0.01: 95% credible set n=5; rs4917;rs4918;rs1900618;rs7635884;rs13073106; rs4917 PIP=0.285539; top-five posterior mass=0.998419.
  W=0.04: 95% credible set n=5; rs4917;rs4918;rs1900618;rs7635884;rs13073106; rs4917 PIP=0.289026; top-five posterior mass=0.998850.
  W=0.09: 95% credible set n=5; rs4917;rs4918;rs1900618;rs7635884;rs13073106; rs4917 PIP=0.289697; top-five posterior mass=0.998919.
  W=0.16: 95% credible set n=5; rs4917;rs4918;rs1900618;rs7635884;rs13073106; rs4917 PIP=0.289934; top-five posterior mass=0.998942.
  W=0.25: 95% credible set n=5; rs4917;rs4918;rs1900618;rs7635884;rs13073106; rs4917 PIP=0.290043; top-five posterior mass=0.998952.

Interpretation boundary: the full-locus single-effect summary-statistic sensitivity is strongly consistent with the prior five-variant result and argues against a simple narrow-window artifact. It does not replace LD-aware multi-signal fine-mapping with an independent/cohort-matched reference.