GENS

What GENS is

GENS was conceived as a cautious way to use raw genotype files from consumer DNA tests (23andMe, AncestryDNA, MyHeritage) in training decisions. Each of five planned traits (aerobic capacity, strength and power, tendon and ligament caution, recovery and chronotype, body composition) was to be shown as a percentile with an uncertainty band, a confidence level, an evidence grade and a note on what the result does and does not mean. Sport selection, diagnoses and deterministic claims were ruled out by design.

DNA pipeline: built and missing

The pipeline is specified as eight steps: ingest, QC, harmonisation, ancestry, scoring, calibration, report and audit. Implemented in Python: file parsers, QC (call rate of at least 95%, missingness, heterozygosity), weighted polygenic scoring with an 80% variant-coverage check, conversion to percentiles and an append-only audit log that records a SHA-256 hash of the input file. Harmonisation, ancestry inference and the population reference distributions are still placeholders. No PGS Catalog scoring files are in the repository, and the deployed service called a stub module that returns placeholder scores.

• Built: parsers, QC, scoring, percentiles, audit log
• Placeholder: harmonisation, ancestry, reference data
• Missing: real polygenic scoring files

Phenotype-first redesign

A specification frozen on 16 February 2026 made DNA optional. The main input became a parent survey, optionally combined with a coach survey and scored by explicit rules rather than machine learning in three regulation domains: sensory, attention and emotional. Each domain gets a support level, a confidence badge and a 4-week plan. Under the specification, athletes under 18 take part only through a parent account, and coaches never see a child’s profile. Release v3.0.0, published on 22 February 2026, made this survey-based mode the default.

Status and what a restart needs

A pilot build was deployed on 17–18 February 2026 on Vercel and Supabase (EU region); nothing has been committed since 22 February 2026, and there are no pilot results. A restart would need real scoring files and reference data, the remaining pipeline steps, legal review of the data-protection documents and a full pilot.