How it works
From scattered data to a daily edge.
Indaga runs the same class of open-source tools and public scientific databases that academic labs use — assembled into one engine, made daily, and kept on your machine.
Step 1 — what you bring
Start with the file you already have.
Tens of millions of people already took a DNA test and never got much from it. Bring that raw file — no new kit, no upload. Add your Watch, glucose and lab panels whenever you like; each one makes the picture sharper.
DNA file
23andMe · MyHeritage · AncestryDNA · VCF
Wearable
Apple Watch · sleep · HR · HRV
Glucose
Dexcom / CGM sessions
Blood panels
lab PDFs, structured
- 23andMeraw data (.txt / .zip)
- MyHeritageraw DNA export
- AncestryDNAraw DNA export
- VCF / gVCFsequencing output
Step 2 — what happens to your DNA
From a raw file to an honest genome.
Your DNA file is aligned, reconstructed into a full genome, indexed and interpreted with the same open databases and methods clinical labs use — all on your own machine. Every step is built to know the difference between “benign” and “we never measured it.”
At a glance
Every step runs on your machine
Your genome · start to answer
- 01
What you bring
Your DNA file
A 23andMe / MyHeritage / AncestryDNA export — the file you already have. Nothing is uploaded; it stays on your machine.
- 02
Align
Matched to the modern genome
Your file is lifted onto the current reference build so it lines up with today's clinical databases.
- 03
Impute
Reconstructed into a full genome
A chip reads only a fraction of your genome. Statistical imputation against a large reference panel fills the gaps to millions of variants — so the ones that matter are actually present — and it runs on your own device.
- 04
Index
Turned into a queryable genome
Every position carries its callability. If a spot was never read, it is marked not measured — unknown, never “you don't have it.”
- 05
Annotate
Interpreted — computed, not looked up
A ClinVar pathogenic / likely-pathogenic screen (position-join), polygenic scores, GWAS trait associations, and a computed ACMG/AMP classification that actually runs the criteria — PVS1 loss-of-function, PM2 / BA1 / BS1 frequency, PP3 / BP4 from ensemble predictors like AlphaMissense and REVEL. Because it is computed, it can classify variants ClinVar has never seen.
ClinVarACMG/AMPPolygenic scoresGWAS - 06
Answer
Queried, and answered honestly
Every answer carries a finding state and a readiness — and refuses to guess about what wasn't measured. Your genetics fuse with your labs, glucose and wearable, so a risk allele is read against your real numbers.
The whole pipeline runs on your machine. Reference databases are downloaded once and queried offline — the only optional network call is a single population-frequency lookup during a variant check, and it is disclosed when it happens.
The model
A high-frequency loop on a deep foundation.
Daily loop · high-frequency
Healthlake foundation · low-frequency
The four stages
Connect, pulse, synthesize, optimize.
- 01
Connect & ingest
Bring the DNA file you already have, and connect your Watch, CGM and blood panels. Everything is structured into your private Healthlake — on your machine.
- 02
Daily pulse
Each morning your sleep and activity update your Biological Clock and locate your Biological Midnight, using a validated 2025 cosinor heart-rate method.
- 03
Deep synthesis
A multi-agent loop — a drafter, an adversarial reviewer that hunts overstated claims, and a reconciler — reads your DNA and history to unlock the “why”.
- 04
Optimize
You get today’s levers, the contextual glucose view, an ask-anything explorer, and a doctor-ready letter for your next visit. Every claim traces to a named source.
Every claim traces back to a named public source — and to a record in your own Healthlake. If you want to check our work, you can.
Indaga is in active development.
We’re building it in the open — privacy-first, cited, and honest about what it can and can’t see.