Build a pedigree by describing your family in plain English

What the GPT is, and what it is not

The Evagene Pedigree Builder is a ChatGPT Custom GPT that takes a natural-language family description and returns a valid Evagene JSON pedigree file. That is its entire job. It does not sequence anything, it does not run risk models, it does not access the Evagene database, and it does not diagnose or advise. The correct framing is ChatGPT helps you produce a valid Evagene file, not “Evagene is inside ChatGPT”.

The file it produces is an ordinary JSON file with the same structure as a pedigree you drew directly in Evagene. When you import it at evagene.net, the platform renders the pedigree using standard ISCN / NSGC notation, attaches any diseases the GPT picked up, marks the proband, and opens the pedigree for analysis. All clinical interpretation — risk models, AI-powered reports, CanRisk / BOADICEA export, inheritance-pattern inference — runs inside Evagene after the file is loaded, not inside ChatGPT.

This is the pattern we call Route A: the GPT produces a file; a human loads the file. A Route B variant using live Actions against the Evagene API is planned — the GPT would create the pedigree on evagene.net directly and return a link — but it is not yet built. Until then, the download-and-load pattern is the whole story.

The full walkthrough

1. Open the Custom GPT

The live GPT is at chatgpt.com/g/g-69ea28b554b8819191ae0f8fd2406d52-evagene-pedigree-builder. Opening the link requires a signed-in ChatGPT account on a paid tier — Plus, Team, or Enterprise. This is OpenAI's requirement for Custom GPTs, not an Evagene paywall. Free-tier ChatGPT users will be prompted to upgrade when they open the link.

Once the GPT opens, you will see four suggested conversation starters covering common consultation scenarios. You can click one, or simply start typing your own description.

2. Describe the family

Tell the GPT about your relatives in whatever order feels natural. What works well:

• Who the pedigree is about — name this person explicitly as the proband if you know the term, or simply say something like “the pedigree is for me” or “this is for my patient”.
• Each relative's relationship to the proband — maternal grandmother, paternal uncle, sister, husband. The GPT is careful to attribute people to the correct side of the family.
• Sex and approximate age — or age at death if deceased.
• Any diseases diagnosed, with approximate ages at onset if known.
• Any genetic test results — positive BRCA1, carrier of cystic fibrosis, variant of uncertain significance in MLH1.
• Ancestry, if relevant — particularly if there is an Ashkenazi Jewish or other population-specific factor in play.
• Consanguinity, if any — “my parents are first cousins” is enough for the GPT to mark the relationship correctly.
• Twin status, adoption, pregnancy losses — the GPT has notation for all of these.

You do not need to know clinical terminology. “My aunt died of breast cancer at 47” is a fine input; the GPT will map it to a diagnosis with an approximate date range and mark the individual as deceased.

3. Answer the clarifying questions

Where the description is ambiguous, the GPT asks before guessing. Common ambiguities it catches:

• Which proband. If you describe a family without naming the index case, the GPT asks who the pedigree is for.
• Which side of the family. If you mention “my grandmother” without specifying maternal or paternal, it asks.
• Disease status when implied but not stated. If you mention a condition without saying who has it, the GPT asks who is affected, who is a carrier, and who has been tested and found clear.
• Biological sex where language is ambiguous. The GPT will not assume from names alone.

The behaviour is deliberate: a silently wrong pedigree is worse than a question. The built-in instructions tell the GPT never to invent a disease or a death you did not state, and never to silently omit a relative you mentioned — even a brief stub entry is preferable to silent omission.

4. Download the JSON file

When the family description is complete, the GPT uses its Code Interpreter to generate the full JSON file — every UUID, every coordinate, every required schema field — and offers it as a downloadable attachment. The file is plain JSON; you can open it in a text editor to inspect it if you want.

Alongside the file, the GPT produces a short summary of what it has built — how many individuals, the generation depth, any notable inheritance patterns it spotted — and points you to evagene.com to render and analyse it.

5. Load the pedigree into Evagene

Open Evagene at evagene.net. From the File menu, choose Load pedigree → Load Evagene file, and select the JSON file you downloaded. The pedigree opens with:

• Individuals rendered as standard ISCN / NSGC symbols — squares for males, circles for females, diamonds for unknown sex.
• Affected relatives filled; carriers with the standard central dot; the proband with a directional arrow.
• Deceased individuals marked with the diagonal slash; dates and events captured.
• Consanguineous unions drawn with a double relationship line.
• Disease sectors colour-coded according to the catalogue entries the GPT created.
• Any genetic test results attached to the correct individuals.

From there the full Evagene workflow is available: add more detail, run BRCAPRO / MMRpro / PancPRO, run the Mendelian inheritance calculators, generate AI-powered clinical interpretation, or export a CanRisk / BOADICEA pedigree file for canrisk.org.

What the GPT handles well

Because the underlying instructions draw on the full Evagene schema, the GPT has a deliberately broad vocabulary. It copes comfortably with:

• Multi-generation structure — three, four, five generations on either side of the proband.
• Mendelian and adjacent inheritance patterns — autosomal dominant, autosomal recessive, X-linked recessive, X-linked dominant, mitochondrial, and the imprinting / UPD category with its parent-of-origin rule (Prader-Willi on the paternal line, Angelman on the maternal line at the same locus).
• Consanguinity — first cousins, double first cousins, uncle-niece, with the appropriate Wright coefficient captured.
• Twins — monozygotic and dizygotic with the correct egg structure.
• Adoption and fostering — the individual is linked to the biological relationship but flagged as adopted or fostered.
• Pregnancy losses and stillbirths — recorded with the appropriate symbol and event.
• Genetic test results — positive, negative, variant of uncertain significance, untested — per individual per gene.
• Approximate dates — “born around 1970” becomes a widened date range rather than a false precise date.
• Cancer detail — primary site, laterality, and basic tumour properties where the description supplies them.

What the GPT does not do

When the GPT is the right choice

Three scenarios where the GPT shortens the path considerably:

Patient-led family history capture. A patient with a complex family can describe it in natural language to the GPT before a genetics appointment. The clinician then loads a structured pedigree and spends the session on interpretation rather than data entry.

Triage at the GP level. A general practitioner who wants to capture family history for a referral letter can build a quick pedigree in a few minutes of conversation, export to PDF from Evagene, and attach to the referral — a structured alternative to free-text family-history notes.

Learning and teaching. Students and educators can iterate on example pedigrees conversationally — build a four-generation Huntington's family, then ask for a variation with reduced penetrance, then a version with adoption — and load each into Evagene to see the standard notation rendering and run the inheritance model.

When drawing directly in Evagene is the right choice

Three scenarios where the GPT is not the fastest path:

• Live clinical consultation. A genetic counsellor taking a family history from a patient across a desk is usually faster with gesture drawing on an iPad. The GPT adds a round-trip through ChatGPT that does not pay off when the person describing the family is already in the room.
• GEDCOM or 23andMe imports. If the starting point is an existing genealogy file or a consumer genomics export, the dedicated GEDCOM and 23andMe importers carry structure the GPT cannot re-derive from prose.
• Highly detailed tumour or test data. For multi-primary cancers, precise receptor-status / grade / stage records, and large gene-panel test results, structured entry in Evagene's properties panel is more reliable than dictation.

Privacy and clinical safety

Conversations with ChatGPT are stored on OpenAI's infrastructure under OpenAI's terms. If you are describing patient data, follow your institution's policy on sending identifiable information to third-party LLMs. In practice, we recommend describing families in de-identified terms — roles and relationships rather than full names, approximate years rather than exact dates of birth — and then filling in identifying detail in Evagene after import, where the data sits under Evagene's data-governance controls.

Evagene's privacy policy covers how the imported pedigree is handled on the Evagene side. The GPT itself is covered by OpenAI's terms; Evagene does not receive the conversation text, only the JSON file you subsequently upload.

Try a starter prompt

If you would like a template to start from, any of these work well:

• “My grandmother had breast cancer at 52, my mother had breast cancer at 45, and my aunt on my mother's side has ovarian cancer at 60. Build a pedigree with me as the proband, 38, female.”
• “I need a pedigree for a cystic fibrosis family — both parents are carriers, two of three children are affected, one is unaffected. The affected children are a boy (6) and a girl (3); the unaffected child is a girl (9).”
• “Build a four-generation Huntington's disease family with autosomal dominant inheritance. The proband is presymptomatic, age 28, tested positive. Affected relatives on the paternal side back to the great-grandfather.”
• “My parents are first cousins and I want to map our family for genetic counselling. I am female, 30, healthy. My older brother is deaf from childhood. My mother's father and my father's mother are siblings.”

Frequently asked questions

Related

• Pedigree drawing software — gesture drawing, keyboard shortcuts, and the full Evagene drawing surface.
• 23andMe pedigree import — consumer-genomics data as another path into the pedigree.
• GEDCOM pedigree software — genealogy-format import and export.
• Pedigree OCR — importing a hand-drawn or printed pedigree from an image.
• Hereditary cancer risk assessment — BRCAPRO, MMRpro, PancPRO, and the cancer family-history scoring models that run on the imported pedigree.
• Mendelian inheritance calculator — the AD / AR / XLR engine plus the non-classical models.
• Evagene platform — API, webhooks, MCP server, and embeddable viewer for teams who want deeper integration than Route A.