Blood manifest: 38 samples
Loaded 17,950,668 rows from 38 blood samples in 20.5s
GC reference: 472,386 probes
  Sex chrom homology mask: 1494 probes
  Pass 1: 4-feat blood GAM (16 workers)...
    20/38 blood samples, 24s, R²=0.5523
    38/38 blood samples, 41s, R²=0.5171
    Sex: 19 female, 19 male, 0 unknown
/usr/local/bin/build_norm_reference.py:247: RuntimeWarning: All-NaN slice encountered
  probe_coverage_median = np.nanmedian(blood_matrix, axis=1)
/usr/local/lib/python3.12/site-packages/numpy/lib/_nanfunctions_impl.py:1593: RuntimeWarning: All-NaN slice encountered
  return fnb._ureduce(a,
    Probe coverage: 471,741/472,386 valid probes
  Preliminary 5-feat GAM on blood (16 workers)...
    20/38 blood prelim 5-feat, 36s, R²=0.9344
    38/38 blood prelim 5-feat, 52s, R²=0.9328
/usr/local/bin/build_norm_reference.py:335: RuntimeWarning: All-NaN slice encountered
  prelim_median = np.nanmedian(prelim_matrix, axis=1)
/usr/local/lib/python3.12/site-packages/numpy/lib/_nanfunctions_impl.py:1593: RuntimeWarning: All-NaN slice encountered
  return fnb._ureduce(a,
    Probe mask: 11,743 masked (2.5%) — 6,142 low cov, 6,524 high var, 3,047 extreme
Wrote batch_reference.parquet (472,386 probes) to ./batch_reference.parquet
Wrote sex_calls.csv (38 samples) to ./sex_calls.csv
Wrote feature_stats.csv (38 samples) to ./feature_stats.csv
Build reference complete in 203.4s total