Open source · macOS · 100% offline
DNA Engine
A genomics workbench that reads your raw-DNA export entirely onyour own Mac and turns it into a curated, citation-backed research readout — marker interpretation, local ancestry, a conservative pharmacogenomics surface, genome quality control, and an encrypted local workspace. No cloud, no account, no telemetry. Your genome never leaves the machine.
Free and open source under Apache-2.0 · build from source on macOS 14+.
What it does
DNA Engine is deliberately conservative: it shows both what a consumer DNA array can suggest and, just as clearly, what it cannot.
Bring your own DNA
Import your consumer raw-DNA export (AncestryDNA / 23andMe-style TSV, CSV, VCF, gzip, or a single-file ZIP). Files are parsed in memory with hardened decompression limits, and every import produces a visible receipt of exactly what was read, accepted, and rejected.
A 403-marker interpretation catalog
Curated markers across roughly fifteen biological domains — methylation, heart & lipids, sleep & circadian, mind & neuro, fitness, metabolism, vitamins, immunity, histamine & gut, longevity, and physical traits — each with an evidence tier, effect text, cross-ancestry transferability, and at least one citation pointer.
Local ancestry, no upload
A bundled, provenance-stamped 2,621-marker reference panel (1000 Genomes phase 3 + gnomAD v4) drives a two-tier admixture model with seeded uncertainty intervals and explicit refusal states. Ancestry is framed as a statistical signal against reference panels — never a passport, nationality, or identity.
Conservative pharmacogenomics
A 14-gene PGx surface reports only per-locus observations and published drug–gene source pairs. It deliberately does not call star alleles, diplotypes, metabolizer phenotypes, medication response, doses, or clinical actions.
QC, panels, ROH & comparison
Genome quality control (build detection, call rate, chip fingerprinting, strand status), fifteen coverage-gated polygenic-tendency panels, runs-of-homozygosity, and a two-file raw-genotype concordance comparison — concordance only, never kinship or relationship inference.
Encrypted local workspace
Your imported genome is sealed in an AES-256-GCM vault whose key is a this-device-only, non-synchronizing Keychain item. There is also a memory-only private session, a deterministic synthetic demo, an inactivity auto-clear, and a verified-erase that confirms deletion before releasing the key.
Your genome never leaves your Mac
There is no network code in the app — no accounts, no telemetry, no analytics, no updater. The original bytes you import are sealed in an AES-256-GCM workspace whose key lives in a this-device-only Keychain item; there is no cloud escrow and no password recovery, by design. A memory-only private session never writes your genome to disk, and verified erase deletes everything and confirms the files are gone before releasing the key. Prefer to look first? A deterministic synthetic demo walks every screen without loading real DNA.
Research questions, not prescriptions
DNA Engine's design rule is fail closed and under-claim. Palindromic sites are reported as strand-unresolved rather than guessed. Pharmacogenomics stays observational. Comparison reports concordance only — never identity or kinship. Lifestyle and pathway surfaces ask neutral research questions instead of handing you a schedule or a supplement to buy. Ancestry carries its own uncertainty and refusal states.
It is built for curious individuals who already downloaded their raw data, students and educators in human genetics, and tinkerers who want a well-documented, dependency-light Swift + Python codebase to extend.
