Standard de-identification was designed for tabular health data, not genomes: Safe Harbor — removes 18 identifiers Expert Determination — certifies low re-identification risk Because the sequence itself re-identifies, you generally cannot treat a genome as de-identified by stripping metadata.
Practical implications: keep the sequence on-device; if you must process variants server-side, use only the specific clinically-relevant variants needed, under a BAA, with consent — and document the residual re-identification risk.