With the one-sided trim in place the robust-drop fractions are much smaller, but a bare >15% drop count still does not separate "expected low-fraction trimming" from a real CNV/LoH/genotyping problem. Options: tie the threshold to the host fraction, or report the dominant reason for the drops (direction of trimmed residuals, clustering on a chromosome arm) instead of a count. Re-measure the post-fix drop distribution before changing the threshold.
Check overlap first. The general REVIEW-sensitivity fix (effect-size / clinical_gating gating, #40) may already cover part of this. Confirm what remains on the robust-drop count specifically before acting.
Low priority
With the one-sided trim in place the robust-drop fractions are much smaller, but a bare >15% drop count still does not separate "expected low-fraction trimming" from a real CNV/LoH/genotyping problem. Options: tie the threshold to the host fraction, or report the dominant reason for the drops (direction of trimmed residuals, clustering on a chromosome arm) instead of a count. Re-measure the post-fix drop distribution before changing the threshold.
Check overlap first. The general REVIEW-sensitivity fix (effect-size /
clinical_gatinggating, #40) may already cover part of this. Confirm what remains on the robust-drop count specifically before acting.Low priority