Everything Evagene does

1. Pedigree drawing & standard notation

Pedigrees are created by drawing directly on an HTML5 canvas. Freehand-draw a circle for a female, a square for a male, or a diamond for unknown sex. Draw a line between two individuals to create a relationship, and draw from a relationship down to an individual to establish parentage. Twin groups are created by drawing chevrons between siblings, with a horizontal bar for monozygotic twins. Keyboard shortcuts accelerate the common actions: m/f/u to add individuals, p for partner, d for the disease palette, n for notes, Delete to remove, Ctrl+A to select all, and F2 to rename.

Symbols render using internationally recognised conventions (NSGC 1995 / 2008 / 2022):

• Base shapes — circle (female), square (male), diamond (unknown), triangle (miscarriage).
• Affection fill — solid fill for affected, dot pattern for carrier, half-fill for heterozygous.
• Disease sectors — colour-coded pie sectors showing per-disease affection status.
• Mortality overlays — diagonal slash for deceased, X for stillborn.
• Proband arrow — configurable compass-direction arrow identifying the index case.
• Relationship lines — double lines for consanguineous unions, annotated with the kinship coefficient.
• Adopted / fostered — square brackets around the individual.

Every panel in Evagene is draggable and resizable from any edge or corner, with a dock icon that highlights when the panel is open. See the pedigree drawing tool, NSGC notation, and symbols reference guides.

2. Guided family-history intake (questionnaire)

For users who would rather answer questions than draw, a guided, step-by-step questionnaire captures a personal and family health history and turns it into a pedigree. It asks about the subject and their relatives — mother, father, siblings, partner, children, grandparents, maternal and paternal aunts/uncles, half-siblings, first cousins, and grandchildren — using Yes/No branch questions so only relevant sections appear. For each person it records name or initials, sex, living status, age (current or at death), conditions from the disease catalogue, genetic test results, and other test results.

On finishing, Evagene creates the individuals, builds the relationships, sets the subject as proband, records the findings, and synthesises any connecting relatives needed to join the tree — either starting a new pedigree or adding relatives to the one already open. It is data capture only: it runs no risk model, produces no risk number, and stores conditions as catalogue references for documentation, not as diagnoses. The same questionnaire is exposed to AI agents through the MCP server (see the platform section).

3. Data capture & reference catalogues

Properties editor

Clicking any individual, relationship, or offspring opens a floating editor:

• Identity — display name, given/family names, prefix/suffix, biological sex (8 options).
• Clinical status — death status (12 options), affection status (11 options), fertility status.
• Physical & lifestyle — height, weight, smoking type/quantity, alcohol units per week.
• Contact — phone, email, address, organisation, title, notes (vCard-compatible).
• Consent — a consent-to-share flag for data governance.
• Relationships — consanguinity coefficient and events (marriage, divorce, separation, partnership).

Disease management (230+ diseases)

A catalogue of 230+ diseases spanning oncology, cardiology, neurology, metabolic disorders, and 20+ multifactorial / polygenic / oligogenic complex conditions. Every entry is cross-referenced to five ontology systems — ICD-10, OMIM, MONDO, Orphanet (ORPHA), and the Disease Ontology (DOID) — as CURIE identifiers, so a disease matches whichever coding system a learner or research dataset uses. A management dashboard supports search, browse-by-taxonomy, quick-add presets (heritable cancers, cardiovascular), inheritance-class filtering, and custom colours. Each entry carries inheritance pattern; genetic parameters (penetrance, age-dependent penetrance tables, homozygous lethality, anticipation, germline mosaicism rate, de novo rate, ancestry-specific carrier frequencies, biological fitness); for complex conditions heritability (h²), prevalence (K), sex-specific prevalence, severity modifiers, and empirical recurrence-risk tables; and marker associations for linkage analysis. Diseases are assigned to individuals with per-disease affection status, laterality, tumour site and properties (ER/PR/HER2, grade, stage), and manifestation tracking with date ranges and certainty levels.

Allergies & traits

• 50+ allergies across six categories with severity, IgE-mediated classification, reaction type, cross-reactivity lists, and prevalence data.
• 50+ traits across six categories with possible values, units, inheritance patterns, heritability estimates, and expression levels.
• Automatic trait inference from 23andMe SNP data — ABO blood type (rs505922, rs8176746, rs8176747), Rh factor (rs590787, rs676785), and FUT2 secretor status (rs602662), shown as visual cards with SNP detail and manually overridable.

Genetic data

• Genetic markers — allele pairs, zygosity, chromosome band positions, mapped to the species genome; editable through the genetics panel.
• Genetic tests — gene name, result (positive / negative / VUS / not tested), method, and date.
• Karyogram viewer — interactive ideogram of all chromosomes with markers plotted at genomic positions, hover tooltips (gene, variant, rsID, genotype), click-to-pin info panels, and embeddable on the canvas. See the karyogram viewer guide.
• Karyotype editor — quick presets (46,XX; 46,XY; 47,XXY; 47,XXX; 45,X; 47,XX,+21; 47,XY,+21).
• Blade navigation — deep genome exploration: Genome › Species › Chromosome › Marker › Disease.
• Ancestry proportions — a hierarchical ancestry catalogue with per-individual proportions and auto-calculation.

Clinical tests & treatments

• Clinical tests — test types with reference ranges, LOINC codes, specimen types, and results.
• Treatments — medication and intervention tracking with dose, frequency, route, date ranges, disease association, prophylactic flag, and status.

4. HPO phenotyping & phenotype-to-gene

Evagene includes a curated Human Phenotype Ontology (HPO) catalogue and a phenotyping panel for recording structured phenotype features against individuals. It is data entry plus educational reference data — documentation of observed family-history detail, not clinical diagnosis or assessment.

• HPO catalogue — a browsable / searchable catalogue of HPO terms, each with its HPO identifier and definition.
• Phenotype features per individual — attach an HPO term as either an observed or an explicitly excluded feature, each optionally qualified with onset and severity. Recording deliberate absences matters for honest family-history documentation.
• Phenotype-to-gene educational associations — from the recorded HPO features, Evagene surfaces candidate associated genes from published HPO-to-gene reference data. This is a list of candidate associations for exploration only — never a ranked likelihood, probability, diagnosis, or recommendation.

HPO phenotyping and phenotype-to-gene associations diagnose nothing, score nothing, and rank nothing; everything is educational reference data and documentation. See the Phenopackets interoperability guide.

5. Risk analysis — 21 published model algorithms

Evagene implements twenty-one published risk-model algorithms in four families on the same pedigree document, for research, teaching, and exploratory use. Outputs are illustrative educational / research artefacts, not clinical recommendations.

Family 1 — Single-gene & adjacent (built-in)

• Autosomal dominant — reduced penetrance, age-dependent penetrance, anticipation, germline mosaicism, parent-of-origin bias, locus heterogeneity, linkage disequilibrium.
• Autosomal recessive — Hardy-Weinberg, genetic-test adjustment, population-specific carrier frequencies.
• X-linked recessive — hemizygosity, mutation rates, biological fitness.
• X-linked dominant — five sex-differential severity sub-modes (equal; males-worse; male-lethal-reproduces — incontinentia pigmenti, OFD1; male-lethal-no-reproduction — Rett, MECP2, CDKL5; metabolic-interference males-unaffected — CFND/EFNB1, EFMR/PCDH19).
• Mitochondrial (mtDNA) — strict maternal transmission, sex-differential penetrance, heteroplasmy scaling. LHON, MELAS, MERRF, NARP, Leigh, Kearns-Sayre, Pearson.
• Digenic two-locus — classical 25% offspring-affected ratio; both_het, one_het_one_hom, both_hom. Usher type 2, certain retinitis pigmentosa, primary congenital glaucoma.
• Imprinting / uniparental disomy (UPD) — mechanism-weighted recurrence across deletion, UPD, imprinting-centre defect, and point mutation; parent-of-origin rule (paternal 15q11-13 IC defect → Prader-Willi; maternal → Angelman). Also Beckwith-Wiedemann, Silver-Russell, TNDM.

Worked guides: autosomal dominant, autosomal recessive, X-linked recessive, germline mosaicism.

Family 2 — Polygenic / oligogenic / multifactorial (built-in)

A liability-threshold recurrence-risk engine grounded in Carter (1961), Falconer (1965), and Reich/James/Morton (1972). For each counselee it classifies affected relatives by degree (MZ twin / 1st / 2nd / 3rd / 4th), uses empirical recurrence-risk tables (Smith, Carter, Harper) when available, falls back to a Falconer liability-threshold calculation from h² and K, and applies four classical modifiers: proband severity, Carter-effect sex bias, multiple affected close relatives, and parental consanguinity as (1 + 2F·h²). 20+ complex conditions are catalogued with heritability and recurrence-risk tables — cleft lip ± palate, Hirschsprung's (oligogenic), type 1 & 2 diabetes, late-onset Alzheimer's, Parkinson's, schizophrenia, bipolar disorder, major depression, epilepsy, pyloric stenosis, neural tube defects, congenital heart disease, and more. See the complex-disease engine guide.

Family 3 — Cancer family-history (built-in + BayesMendel R sidecar)

BayesMendel Bayesian models via an R sidecar: BRCAPRO (BRCA1/2 carrier probability, breast/ovarian risk), MMRpro (Lynch syndrome, MLH1/MSH2/MSH6), and PancPRO (pancreatic). Family-history scoring models (no R sidecar needed):

• Claus (CASH-derived lifetime breast-cancer risk; Claus, Risch & Thompson 1994).
• Couch (logistic BRCA1 pre-test probability; Couch 1997).
• Frank / Myriad (empirical BRCA1+BRCA2 probabilities; Frank 2002).
• Manchester Scoring System (BRCA1/BRCA2 point totals against 10% / 20% cut-offs; Evans 2004).
• NICE CG164 (2013) / NG101 (2018, updated 2023) familial breast-cancer categorisation — near-population / moderate / high, with the matched-trigger list. The categorisation illustrates how the published NICE thresholds may be represented; it is not a referral recommendation issued by the software.
• Amsterdam II criteria (Vasen 1999) and Revised Bethesda Guidelines (Umar 2004) for Lynch syndrome — pedigree-derivable criteria evaluated directly. Where these published guidelines describe testing considerations, Evagene represents the criterion; it does not make that recommendation.
• Gail (NCI BCRAT; Gail 1989 and later revisions) — 5-year and lifetime breast-cancer risk.
• Tyrer-Cuzick (published Tyrer/Duffy/Cuzick 2004 algorithm) — 10-year and lifetime breast-cancer risk including family history, reproductive factors, BMI, mammographic density (BI-RADS 1–4), atypical hyperplasia, LCIS, and Ashkenazi ancestry. Clearly labelled in-app and in reports as an IBIS-style approximation — not the official IBIS Breast Cancer Risk Evaluator binary, whose full coefficients are not publicly released.
• Prostate-cancer family-history scoring (illustrative / educational, gated behind the intended-use acknowledgement) — the twenty-first algorithm: number and degree of affected relatives, ages at diagnosis, bilateral / early-onset features. Where it cites a published threshold it attributes it to the external guideline and states that Evagene does not make that recommendation.

Family 4 — External-tool bridge: CanRisk / BOADICEA export

One-click download of a tab-separated ##CanRisk 2.0 file populated with pedigree structure, ages at diagnosis, BRCA1 / BRCA2 / PALB2 / ATM / CHEK2 statuses, Ashkenazi flag, and reproductive factors (GET /api/pedigrees/{id}/risk/canrisk). BOADICEA is not bundled — the model is licensed by the University of Cambridge; the clinician uploads the exported file at canrisk.org and runs BOADICEA there. The export path is the legally clean route to the NICE-recommended tool. See BOADICEA / CanRisk context.

Supporting risk surfaces

• Population-conditioned carrier priors — for any individual with recorded ancestry, prior carrier probabilities across the AR / XLR catalogue, weighted by ancestry proportions, with every row attributed to an explicit source (published population key, weighted mix, or Hardy-Weinberg estimate). Ancestry-gated by design: no ancestry recorded, no personalised carrier number.
• Reproductive / breast-health fields — age at menarche / menopause / first live birth, parity, hormone-therapy years, prior benign biopsies, atypical hyperplasia, LCIS, mammographic density, and BMI inputs for Gail, Tyrer-Cuzick, and CanRisk.
• Germline mosaicism posterior probability — computed from pedigree structure, with optional somatic variant allele fraction (VAF) input from blood or sperm sequencing and joint-parent logic that exonerates one parent when the other is confirmed mosaic. Parameterised per condition by de_novo_rate, germline_mosaicism_rate, and germline_mosaicism_sex_bias. See the germline mosaicism calculator.
• Outputs & reports — carrier and genotype probabilities, future risk by age, per-member breakdown, joint outcome probabilities, LOD scores, flanking marker information, recurrence risks as "1 in X", thresholds met, and a risk karyogram of elevated-risk loci.
• Differential-diagnosis mode — analyses affected siblings from unaffected parents, generating weighted hypotheses (reduced penetrance, gonadal mosaicism, new mutation, autosomal recessive) with posterior probabilities.
• Batch risk exploration — screens a proband across all diseases at once with configurable thresholds and population selection, producing four educational report types (plain-language, structured, carrier-probability, and AI-assisted draft summary). See batch risk exploration.

6. Related concepts (educational correlation graph)

A hand-curated correlation graph of educational associations between the entities Evagene already knows — diseases, traits, clinical-test results, allergies, and genetic markers. It is reference data in the same family as the catalogues, not a risk-analysis tool and not a claim about any person.

• Edges carry a relationship from a fixed eight-term vocabulary (genetic_association, shared_gene, biomarker, associated_finding, associated_condition, shares_features, sequela, risk_association), a short neutral note, an optional strength, and a direction flag.
• Two sources — over 1,100 hand-authored curated edges, plus genetic edges auto-derived from the marker catalogue at start-up.
• Status-qualified clinical-test edges — a low result surfaces different associations than a high one (low ferritin → iron-deficiency anaemia; high ferritin → hereditary haemochromatosis).
• UI & access — the Related concepts panel, the open /api/correlations endpoints, and the MCP related_concepts tool (always carrying the educational-use disclaimer).

It does not diagnose, produces no risk number, runs no inheritance analysis, and asserts nothing about any individual. Explore the Related Concepts Explorer, biomarker associations, and gene associations.

7. Consanguinity, kinship & IBD

Wright's path-coefficient method calculates consanguinity coefficients for all relationships, displayed on relationship lines, cached for performance, and fed into risk calculations. A dedicated relatedness / IBD analysis module extends this for educational and research exploration of pedigree structure:

• Full pairwise kinship coefficients across every pair of individuals.
• Per-individual inbreeding coefficients.
• Jacquard condensed identity coefficients and the expected genome-wide IBD sharing proportions k0 / k1 / k2 for each pair.
• A gene-dropping simulation that drops founder alleles through the pedigree to illustrate empirically how relatedness and IBD sharing arise from the family structure.

This is an educational and research tool for studying pedigree relatedness from documented family structure. It is not a forensic, identity-testing, parentage-testing, or relationship-verification tool. See the consanguinity calculator.

8. AI draft summaries, descriptions, notes & search

• AI-assisted draft summaries for educational / research review — Evagene generates draft educational summaries of a pedigree (structural observations, family-structure notes and documentation gaps, data-completeness prompts, and discussion prompts linked to published literature). These are drafts for human review only — not clinical advice, not diagnostic output, and not intended to inform clinical decisions. Summaries are placed on the canvas as anchored notes attached to the proband, and a quota system tracks usage.
• Pedigree description engine — a built-in generator producing structured English-language descriptions of the pedigree and proband (family structure, genetic findings, shared carrier statuses, inheritance considerations, data gaps).
• Anchored canvas notes — floating notes anchored to any individual, following them when moved; used for AI reports, descriptions, and karyogram charts.
• Advanced search — a query builder filtering individuals by sex, affection status, death status, age range, genetic and clinical test results, treatments, and events, with relational scope (parents, siblings, children, ancestors, descendants) and AND/OR match modes.

9. Data exchange & interoperability

Eleven formats move data in and out of Evagene:

The Evagene Pedigree Builder (ChatGPT Custom GPT) lets a user describe a family in plain English; the GPT asks clarifying questions, then produces a downloadable Evagene-compatible JSON file that the user loads into Evagene. It is strictly a data-capture aid — it does not give medical advice, diagnose, or run risk models. Related guides: GEDCOM converter, 23andMe import, Phenopackets, pedigree OCR.

10. Platform & integrations

A platform layer turns pedigree data into infrastructure for external systems and AI agents. Full reference: Platform & Integrations, REST API, MCP server, embeddable viewer.

REST API keys

Long-lived, scoped, rate-limited keys for the full REST API. Format evg_<43 chars> (256-bit entropy), SHA-256 hashed at rest. Three scopes (read, write, analyse), configurable per-minute (default 60) and per-day (default 10,000) limits with sliding-window enforcement, up to 20 keys per user, optional expiry. Auth via X-API-Key header or ?api_key=.

Bring Your Own Key (BYOK) LLM

Users supply their own Anthropic (Claude, default claude-sonnet-4-20250514) or OpenAI (GPT, default gpt-4o) key for all AI interpretation and template execution. Keys are encrypted at rest with Fernet (AES-128-CBC + HMAC-SHA256), keyed from the server's JWT secret. When configured, BYOK bypasses the standard rolling 24-hour quota and the default model can be overridden. See LLM pedigree analysis.

Webhooks

HMAC-SHA256-signed HTTP POST notifications on data change. Eight event types (pedigree.created/updated/deleted, individual.created/updated/deleted, analysis.completed, import.completed), unique per-webhook secrets, signature headers, up to 20 webhooks per user, and delivery history of the last 20 attempts each.

Analysis templates

Reusable custom AI prompt templates with variable injection ({{pedigree_description}}, {{proband_name}}, {{proband_sex}}, {{disease_list}}, {{risk_summary}}), a system prompt and a user-prompt template, optional public sharing, and up to 50 per user.

MCP server — 27 tools for AI agents

A stdio-mode Model Context Protocol server exposes 27 tools to AI agents, authenticated via the EVAGENE_API_KEY environment variable and connecting directly to the database (no HTTP round-trip). Compatible with Claude Desktop, Claude Code, and any MCP-capable client; changes are immediately visible in the web UI.

Embeddable pedigree viewer

Drop-in pedigree diagrams for third-party sites via three methods: a full HTML page via iframe (/api/embed/{id}), raw SVG (/api/embed/{id}/svg), and a JavaScript snippet (/api/embed/snippet.js) that auto-discovers data-evagene-pedigree attributes. Standard notation, API-key or public-pedigree authentication, responsive SVG with viewBox scaling. A complementary free drawing iframe embeds the full editable canvas in any course page, LMS module, or blog.

11. Open-source integration examples

Evagene maintains a free, MIT-licensed public repository of 25 worked integration examples at github.com/xiasma/evagene-integration-examples, spanning Python, Node.js / TypeScript, .NET / C#, R, and Go. Each demo leads with user value, pins its dependencies, splits code by responsibility, and ships with unit tests. A shared/ directory holds reusable API-client utilities, and getting-started.md walks through first run.

API & intake patterns

• nice-traffic-light — represents the NICE CG164/NG101 familial-breast-cancer categorisation (near-population / moderate / high) from a pedigree as a coloured indicator. (Python, Node, .NET, R · REST API)
• family-history-intake-form — a web form capturing three generations and turning them into a structured pedigree. (Python, Node, .NET · REST API)
• chat-bot — a Slack / Teams bot that surfaces pedigree details and the NICE category. (Node · REST API)
• pedigree-puzzle-generator — random inheritance-pattern teaching quizzes (AD / AR / XLR / XLD / MT) with answer keys. (Python, Node · REST API)

Geneticist & research tooling

• canrisk-bridge — exports a pedigree to the BOADICEA / canrisk.org file in one command. (Python, Node, .NET, R · REST API + file formats)
• bayesmendel-comparator — runs BRCAPRO / MMRpro / PancPRO side by side for comparison. (R, Python · REST API + R sidecar)
• tumour-board-briefing — compiles the family-history scoring model outputs (Claus, Couch, Frank, Manchester, NICE, Tyrer-Cuzick) for a pedigree into a one-command PDF, for educational / research review — not a clinical report. (Python · REST API + SVG)
• archive-triage-runner — batch-annotates a folder of GEDCOM files with each pedigree's NICE category representation in a CSV; a data-annotation example, not automated triage or clinical flagging. (Python, .NET · REST API)
• cascade-screening-letters — generates draft letter templates from a pedigree using analysis templates; illustrative scaffolding only, making no recommendation about who should be contacted or tested. (Python, Node · REST API + templates)
• pedigree-diff — compares two pedigrees and produces a text / JSON / Markdown changelog. (Python · REST API)
• couple-carrier-risk — combines two 23andMe exports into a carrier-frequency and offspring-risk table (AR + XLR). (Python, R · REST API + file formats)
• publication-figure-renderer — print-quality SVG with optional de-identification (I-1, II-3 labels). (R, Python · SVG)
• research-cohort-anonymiser — strips identifiers, preserves structure, and estimates k-anonymity. (Python, R · REST API)
• notebook-explorer — Jupyter (Python) and Quarto (R) notebooks for interactive risk what-if exploration. (Python, R · REST API)

AI agents & multimodal input

• mcp-server — a Model Context Protocol server for Claude Desktop / Cursor, in three languages. (Python, Node, Go · MCP)
• call-notes-to-pedigree — turns a counselling transcript into a structured pedigree via Claude extraction. (Python, Node · BYOK LLM)
• pedigree-ocr — a hand-drawn pedigree photo to an Evagene pedigree via Claude Vision. (Python · BYOK LLM)
• voice-driven-intake — audio session → Whisper transcript → Claude extraction → pedigree. (Python · BYOK LLM)

Integration, workflow & embeds

• fhir-bridge — maps pedigrees to and from FHIR R5 FamilyMemberHistory Bundles: a data-format interoperability example showing how the open export formats bridge to other data standards, not EHR or clinical-workflow integration. (Node, .NET · REST API + file formats) See the HL7 FHIR pedigree guide.
• webhook-audit-blotter — an HMAC-verified webhook receiver with a tamper-evident SQLite log. (Node, .NET, Python · webhooks)
• clinic-referral-dashboard — an illustrative developer UI combining webhook events, Server-Sent Events, and embedded pedigree cards; reference code, not a clinical monitoring or decision-support dashboard. (Node · webhooks + embeddable viewer)
• longitudinal-risk-monitor — a webhook-automation example: a scheduled job that posts to Slack when a pedigree's NICE category representation changes between runs. (Python · webhooks + REST API)
• xeg-upgrader — converts legacy v1 .xeg XML into a current pedigree. (.NET, Python · REST API + file formats)
• shareable-pedigree-link — mints a scoped read-only key and returns an iframe-ready embed URL. (Node, Python · REST API + embeddable viewer)
• browser-extension — a Manifest V3 extension that highlights pedigree UUIDs with a side-panel viewer. (Node / WebExtensions · embeddable viewer)

12. Help, documentation & learning

• 1,900+ help guides across six cross-linked catalogues — diseases (230+), traits, allergies, clinical tests (LOINC), markers / genes (NCBI Gene, OMIM, ClinVar), and treatments (RxNorm, BNF, DrugBank) — at evagene.net/help.
• Educational guides — introductory through advanced genetics, plus clinical-genetics references for patients and professionals.
• 8 role-specific workflow guides — one per persona, in basic and advanced tiers.
• A 46-page educational topic library at MSc / FRCPath level (1,300–2,000 words per page, with inline citations) across 12 major topics — molecular architecture of genes, gene-expression mechanisms, regulation of gene activity, mutation biology, chromosomes & cell division, Mendelian genetics, quantitative & complex traits, population & evolutionary genetics, haemoglobin & development, cancer genetics, metabolic genetics & therapeutics, and clinical practice / ethics — each with two to three subtopic pages.
• Learn Genetics — seven structured reading pathways (Foundations, Molecular biology, Clinical genetics, Cancer genetics, Population & quantitative, Therapeutics & gene therapy, Recent advances) with defined audiences, prerequisites, and time estimates.
• Genetic Current — a daily-updated genetics news section that aggregates roughly sixteen public sources, deduplicates and clusters by cross-source consensus, and produces factual summaries with eight audience indices (researchers, GPs, oncologists, genetic counsellors, educators, students, genealogists, patients), an archive, and full-text search.
• Research & citations — peer-reviewed papers citing Evagene and the U.S. Department of Veterans Affairs Technical Reference Model listing.

13. Who Evagene is for

Evagene is an educational and research tool used for teaching, training, documentation, and exploratory research — not for clinical decision-making. Professional users are addressed as educators, researchers, and learners:

• Research geneticists — Mendelian modelling, linkage analysis with LOD scores, marker management, species flexibility.
• Clinical geneticists (teaching, research, documentation) — accurate standard notation and published-algorithm exploration on anonymised / synthetic pedigrees.
• Genetic counsellors & nurses — plain-language educational summaries, consent tracking, role-specific workflow guides.
• Oncologists — educational exploration of BRCAPRO, MMRpro, and PancPRO on anonymised / synthetic pedigrees.
• General practitioners — three-generation documentation for training and reference.
• Patients & families — gesture drawing, 23andMe import, image import, and plain-language educational summaries (not medical advice).
• Genealogists — GEDCOM 5.5.1 import/export, multi-generation support, automatic consanguinity detection.
• Educators & students — the 1,900+ guide catalogue, the topic library, and standard-notation enforcement.

14. What Evagene does not claim

• Evagene is not a medical device. It is not FDA-cleared, CE-marked, UKCA-marked, or otherwise registered as clinical software.
• Risk-model outputs are illustrative educational / research artefacts, not clinical recommendations, screening triggers, or referral decisions.
• Tyrer-Cuzick output is an IBIS-style approximation of the published Tyrer / Duffy / Cuzick 2004 algorithm — not the official IBIS Breast Cancer Risk Evaluator binary, whose full coefficients are not public.
• BOADICEA is not bundled. The full computation runs off-platform at canrisk.org under University of Cambridge licence, after the user uploads Evagene's CanRisk file export.
• The HPO phenotyping, phenotype-to-gene, related-concepts, and IBD surfaces are educational reference data and documentation — they diagnose nothing, score nothing, and assert nothing about any individual.
• The open-source integration examples are illustrative reference code for developers, not endorsed clinical workflows; Evagene remains a research and education tool.