Skip to content

Validate contamination-flag thresholds on clean controls and other panels #44

Description

@davmlaw

CONTAMINATION_WARN_FRACTION (0.2%), CONTAMINATION_REVIEW_FRACTION (1%), and the empirical p10 floor were all set from SRP434573 alone. The p10-floor estimator assumes a useful fraction of no-carrier sites and a particular allele-frequency spectrum, and even on clean data it reports a non-zero floor (~0.15% at 1% error, 2000x) from binomial sampling spread (shrinks as ~1/sqrt(depth)).

Measure the false-positive rate on genuinely uncontaminated high-depth samples and confirm the empirical floor behaves on panels with different marker allele frequencies. Validation work, not a code change, but it gates trusting any LoD floor built on the contamination estimate in deployment.

Evidence: [speculative]. Low priority (paper first). Was Step 32 in the retired overall plan.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions