UNDERSTAND THE EVIDENCE
Methods, evidence and limitations
A transparent account of what DNAObs checks, what its results mean, and what has not been scientifically validated.
Input quality and reference compatibility
- The upload parser checks supported layouts, bounded decompression, missing calls and conflicting records. Declared provider and genome-build information is not independent verification.
- Variant annotation checks reference alleles against its configured reference and reports database provenance. Eligibility depends on the original VCF and the checks performed on the server.
Marker associations are not disease probabilities
- The association catalog matches observed alleles. Matching an allele does not establish a diagnosis, penetrance, severity or calibrated personal risk.
- An absent catalog allele or an untested marker cannot rule out a condition. Confirm important findings with an appropriate healthcare professional.
- DNAObs does not display composite disease-risk probabilities derived from simple marker counts.
Ancestry and trait limitations
- Reference-population models depend on their training/reference panels and your marker overlap. Country affinities are heuristic comparisons, not evidence of nationality or birthplace.
- Unvalidated phenotype models remain unavailable. Scientific confidence is not inferred from a successful upload or a software test.
Validation status
- Software regression tests check processing, ownership, persistence, payment events and input handling. Small normalization fixtures check representation and reference compatibility.
- These checks do not establish medical, ancestry or phenotype accuracy. Independent held-out validation across representative populations is still required before stronger prediction claims can be made.
Resources & Playbooks
Dive deeper into DNAObs methodology, compliance, and applied genomics workflows.