Platform

The platform for the genome of the species

Pipeline v7 · 16 active analysis modules · 6 live PRS traits

The flywheel · your Mirror over time

Consented exchange
drag to rotate · tap a system

The system · what the platform is made of

Ingestsupported raw DNA23andMe · VCF · FASTQHarmonizebuild · strand · liftoverpipeline v7THE ENGINE16 active modules6 live PRS traitsDBKG · PGx-gatedTHE TWINMirrordigital twin · 3 layersTHE SURFACESMirrorOptimizeLabNetworksOriginsGenomeTHE NETWORK · CONSENT-NATIVEprofile visibilitygenomic opt-inprivateINDIVIDUAL · LINEAGE · SPECIESTHE HERU SUBSTRATEreference data + calibrationLD panels · 5 populationsgnomAD · HGDP + TGPPGS CatalogGRCh37 / 38 referencesCalibration corpus
01Ingest
02Engine
03Twin
04Surfaces
05Substrate
Expand
At scale

The same instrument at every resolution

Most genomics ends at a report. Haeckel keeps supported genomic evidence in one encrypted, queryable model, then exposes it through analysis, conversation, lineage context, consented discovery, and statistical modeling. Each surface has its own data requirements and maturity.

The model

The digital twin has three layers

Each person has an evidence model spanning available genotype, phenotype, and extended context; it is not a complete replica of their biology.

Genotype

The genome. Variants from supported consumer DNA, VCF, and FASTQ files, stored with their harmonized coordinates.

Phenotype

The expressed biology. Health, traits, biomarkers, wearables, hormones, cognition, body composition.

Extended phenotype

What the genes produce beyond the body. Lineage, offspring, behavior, environment, footprint.

The stack

Five operations on the twin

The platform is organized around five operations. Each is explicit about its current maturity, inputs, and limits.

DigitalizeInstrument

Ingest a supported consumer DNA, VCF, or FASTQ file, harmonize its variants, and store the resulting biological model as encrypted, queryable data.

OptimizeDecision workspace

Organize goals, available biological signals, experiments, and supported drug-gene safety checks in one workspace. Daily automation is still partial.

CompareEvidence boundary

Compare consented genomic evidence only where both profiles meet the required coverage. Haeckel does not currently offer offspring or embryo modeling from stored genomic projections.

InteractThe agent

Ask natural-language questions over the evidence available in the encrypted twin and open the relevant visual context.

ExchangeConsented discovery

Profile visibility and genomic matching use separate consent controls. People choose whether their profile can be discovered and whether genomic evidence participates.

Resolution

Individual. Lineage. Species

Comparison and discovery become more informative as the consented reference space grows, while individual evidence and permissions remain attached.

IndividualResolve one genome across every layer
LineageInheritance, and the generations forward and back
SpeciesConsented population comparison and research context
Data flow

From a raw file to a queryable model

Five stages, each independently tested and versioned.

IngestionSupported consumer DNA · VCF · FASTQ
HarmonizationBuild detection · strand · liftover
16 active modulesAncestry, health, traits, pharma, lineage
PRS EngineHarmonized live scoring for 6 traits
MirrorThe conversational interface
Modules

16 active modules on a versioned pipeline

Each active module names its scope, method, coverage requirements, and known limits. Research engines that are not in the live runtime are documented separately.

Ancestry v4.0
MLE-EM + spatial thinning + Bootstrap Wald + AMR deconvolution
Health
ClinVar matching, sex-aware X-linked logic, 76-gene penetrance lookup
Traits
Genotype-to-phenotype across monogenic and polygenic traits
Haplogroup v2.1
Recursive phylogenetic traversal, back-mutations, RC fallback
APOE
Four-tier resolution across chip versions
HLA
Imputation and typing of the MHC region
PCA
Principal-component projection onto reference cohorts
Archaic
Coverage-gated Neanderthal and Denisovan marker estimates
ROH
Runs of homozygosity, consanguinity inference
HIrisPlex
Eye / hair / skin prediction, coverage-gated
Pharmacogenomics
Star-allele calling, CPIC-guided interpretation, HLA and G6PD safety signals
Nutrigenomics
Diet-relevant variant interpretation
Kinship
KING-robust relationship inference
Data Quality
Coverage, call rate, and build integrity gates
Runtime
Next.js 14 · Vercel
Database
Neon Postgres + pgvector
1536-dimensional embeddings
File storage
Cloudflare R2
Encrypted at rest
Reference panels
1000 Genomes + HGDP
62 populations · 5 LD reference panels
Platform — Haeckel