Methodology & Provenance
How a HelixlyAI report is built: source assembly, hashing, evidence grading, and supersession. This page documents the pipeline and contains no personal findings. All values shown are synthetic and for demonstration only.
Provenance
Source & Assembly
A HelixlyAI report is built from a single consumer DNA export. The pipeline aligns every call to the GRCh37/hg19 assembly and builds a genotype-truth INDEX directly from the raw chip file. Each genotype shown on every page resolves to one row in that INDEX, identified by its rsID and chromosome position.
Hashing & Provenance
Every document records the SHA-256 of its source export and the SHA-256 of the generated INDEX; both hashes appear in the provenance block at the top of each page. When a document replaces an earlier version, the prior hash is recorded under Supersedes, so the chain from raw file to rendered claim stays auditable.
Evidence Grading
Every statement carries a significance tier and an evidence grade:
- Tier 1 — variants with established clinical action.
- Tier 2 — variants with moderate or emerging evidence.
- Tier 3 — variants reported for transparency; current evidence is insufficient for clinical action.
CPIC assigns drug–gene guidance Levels A–D, where Level A reflects the strongest prescribing evidence. Each statement ties to an INDEX rsID and the observed genotype, cites an allow-list source, and is written in the present tense.
Allow-list Sources
A HelixlyAI report cites only these eight authoritative sources:
- PubMed ↗ — primary literature.
- ClinVar ↗ — variant clinical significance.
- CPIC ↗ — pharmacogenomic prescribing guidelines.
- PharmGKB ↗ — gene–drug knowledge base.
- FDA PGx ↗ — pharmacogenomic biomarkers in drug labeling.
- GWAS Catalog ↗ — trait–variant associations.
- dbSNP ↗ — variant identifiers and positions.
- gnomAD ↗ — population allele frequencies.