Hereditary Cancer Risk Assessment: Understanding BRCAPRO, MMRpro, and PancPRO

What Is Hereditary Cancer?

Hereditary cancer syndromes are caused by pathogenic variants in tumour suppressor genes or DNA repair genes that are present in every cell of the body from birth. Because these variants are in the germline, they follow Mendelian inheritance patterns and can be transmitted from parent to child with a defined probability—typically 50 percent for autosomal dominant conditions.

The distinction between hereditary and sporadic cancer is clinically critical. Sporadic cancers develop through the stepwise accumulation of somatic mutations in a single cell lineage, usually later in life and without a clear familial pattern. Hereditary cancers, by contrast, tend to present with characteristic features: earlier age of onset, bilateral or multifocal tumours, multiple primary cancers in the same individual, and a recognisable pattern of specific cancer types across generations.

Several high-penetrance genes are central to the most common hereditary cancer syndromes:

• BRCA1 and BRCA2 — Pathogenic variants in these genes confer substantially elevated lifetime risks of breast cancer (up to 70–80%) and ovarian cancer (up to 40–45% for BRCA1, 15–20% for BRCA2). These genes encode proteins essential for homologous recombination repair of double-strand DNA breaks.
• MLH1, MSH2, and MSH6 — These mismatch repair (MMR) genes are responsible for Lynch syndrome, the most common hereditary cause of colorectal cancer. Pathogenic variants increase risk for colorectal cancer (up to 80% lifetime risk), endometrial cancer (up to 60%), and several other malignancies including ovarian, gastric, and urinary tract cancers.
• Other high-penetrance genes — TP53 (Li-Fraumeni syndrome), APC (familial adenomatous polyposis), CDKN2A (familial melanoma and pancreatic cancer), STK11 (Peutz-Jeghers syndrome), and PALB2 (breast and pancreatic cancer) are among the growing list of genes with established clinical actionability.

The inheritance pattern for most of these syndromes is autosomal dominant with incomplete penetrance. This means that carrying one copy of the pathogenic variant is sufficient to increase cancer risk substantially, but not every carrier will develop cancer. Penetrance—the probability of developing the disease given a pathogenic genotype—varies by gene, specific variant, sex, and modifier genes. This variability is precisely why Bayesian risk models, which account for penetrance functions and family structure, are essential for accurate risk estimation.

The Role of Family Pedigrees in Cancer Risk Assessment

The family pedigree is the foundational data structure for hereditary cancer risk assessment. It is far more than a diagram: it is a systematic encoding of biological relationships, phenotypic information, and temporal data that risk models require for their calculations. A well-constructed pedigree allows clinicians and genetic counsellors to visually identify inheritance patterns, flag individuals at elevated risk, and determine which quantitative models to run.

For a pedigree to serve as reliable input for models like BRCAPRO, MMRpro, and PancPRO, it must capture several categories of information:

• Family structure — All first-, second-, and third-degree relatives on both maternal and paternal sides. This includes parents, siblings, children, grandparents, aunts, uncles, and first cousins. The more complete the pedigree, the more informative the risk calculation.
• Cancer diagnoses — Specific cancer types for each affected individual, not simply "cancer." The distinction between, for example, breast and ovarian cancer is critical for BRCAPRO, just as the distinction between colorectal and endometrial cancer matters for MMRpro.
• Ages of diagnosis — Early-onset cancers carry substantially more weight in risk models because they are more likely to reflect an underlying germline variant. A breast cancer diagnosis at age 35 is far more informative than one at age 72.
• Current ages of unaffected relatives — Unaffected relatives contribute to the risk calculation by providing person-years of observation. A family with many unaffected elderly women is less likely to harbour a BRCA variant than a small family where the absence of cancer may simply reflect limited observation.
• Vital status and age at death — For deceased relatives, the age at death and cause of death help models determine the observation window during which cancer could have been diagnosed.
• Prior genetic testing results — If any family members have undergone genetic testing, these results can be incorporated into the model to refine carrier probability estimates.
• Ethnicity and ancestry — Population-specific allele frequencies affect prior probabilities. For example, Ashkenazi Jewish populations have significantly higher carrier frequencies for specific BRCA1 and BRCA2 founder variants.

Professional guidelines from organisations such as the National Comprehensive Cancer Network (NCCN) and the American College of Medical Genetics and Genomics (ACMG) recommend a minimum three-generation pedigree for hereditary cancer evaluation. Importantly, cancer predisposition genes can be inherited from either parent, so both lineages must be evaluated even when the family history appears to cluster on one side.

Pedigree construction has traditionally been a manual process—drawn on paper or assembled in standalone desktop applications. Modern pedigree drawing software has transformed this workflow by providing structured data entry, standard genetic notation, and direct integration with risk models, eliminating the need to manually re-enter family history data into separate calculators.

Understanding BRCAPRO

BRCAPRO is the most widely used and extensively validated Bayesian risk model for estimating the probability that an individual carries a pathogenic variant in BRCA1 or BRCA2. Developed as part of the BayesMendel framework at Johns Hopkins and Harvard, it has been cited in thousands of research publications and is recommended in clinical guidelines worldwide.

What BRCAPRO Calculates

BRCAPRO produces two primary categories of output. First, it estimates the posterior carrier probability: the probability that the counselee carries a pathogenic variant in BRCA1, in BRCA2, or in either gene. These probabilities are independent and sum to less than one (the remainder being the probability of carrying no pathogenic variant in either gene). Second, BRCAPRO calculates cumulative cancer risk projections: the probability of developing breast cancer or ovarian cancer by specific ages (e.g., by age 50, 60, 70, or 80), conditional on the estimated carrier status.

How Bayesian Updating Works

The mathematical foundation of BRCAPRO is Bayes' theorem, which provides a principled framework for updating beliefs in light of new evidence. The model begins with prior probabilities derived from population-specific allele frequencies for BRCA1 and BRCA2 pathogenic variants. For a general European population, the combined carrier frequency is approximately 1 in 400; for Ashkenazi Jewish populations, it is approximately 1 in 40 owing to three common founder variants.

Each relative in the pedigree contributes a likelihood ratio that quantifies how much more (or less) likely their phenotype is if the counselee is a carrier versus a non-carrier. Affected relatives with early-onset breast or ovarian cancer generate high likelihood ratios that increase the posterior carrier probability. Unaffected elderly relatives generate likelihood ratios less than one, pulling the estimate downward by providing evidence against the presence of a high-penetrance variant in the family.

The model uses Mendelian transmission probabilities to propagate genotype information through the pedigree. For each relative, the model considers all possible genotype configurations consistent with the family structure and weights them by their probability. This is computationally achieved through the Elston-Stewart peeling algorithm, which efficiently traverses the pedigree tree without enumerating all possible joint genotype configurations explicitly.

Required Inputs

BRCAPRO requires the following data for each individual in the pedigree: sex, relationship to the counselee, breast cancer status and age of diagnosis (if affected), ovarian cancer status and age of diagnosis (if affected), current age (if alive and unaffected), age at death (if deceased), and any prior BRCA genetic testing results. The model also accepts information about bilateral breast cancer, male breast cancer (a strong indicator of BRCA2), and contralateral breast cancer. Ethnicity is used to select the appropriate prior allele frequencies and penetrance functions.

Clinical Utility

BRCAPRO results directly inform clinical decisions. A carrier probability above a defined threshold (commonly 10% or greater) may trigger a recommendation for genetic testing. The cumulative risk projections guide screening intensity—for example, whether to recommend annual breast MRI in addition to mammography—and inform discussions about risk-reducing mastectomy or oophorectomy. Importantly, BRCAPRO provides a quantitative basis for these decisions, moving beyond qualitative pattern recognition to evidence-based risk stratification.

Understanding MMRpro and Lynch Syndrome Screening

MMRpro extends the BayesMendel framework to address Lynch syndrome, previously known as hereditary non-polyposis colorectal cancer (HNPCC). Lynch syndrome is the most common hereditary predisposition to colorectal cancer and is caused by pathogenic variants in the DNA mismatch repair genes MLH1, MSH2, and MSH6.

Like BRCAPRO, MMRpro is a Bayesian model that computes carrier probabilities and cumulative cancer risks based on family history. However, it addresses a broader spectrum of cancers. Individuals with Lynch syndrome face elevated risks of colorectal cancer (up to 80% lifetime risk depending on the gene), endometrial cancer (up to 60% for MLH1 and MSH2 carriers), and moderate increases in risk for ovarian, gastric, urinary tract, hepatobiliary, small bowel, and brain cancers.

Tumour Marker Integration

A distinguishing feature of MMRpro is its ability to incorporate tumour marker data as intermediate phenotypes. Microsatellite instability (MSI) testing and immunohistochemistry (IHC) for mismatch repair proteins are routinely performed on colorectal and endometrial tumour specimens. These results provide direct evidence about whether a tumour's mismatch repair machinery is intact. MMRpro uses this information as an additional likelihood layer, substantially refining the carrier probability estimate beyond what family history alone can achieve.

For example, a colorectal cancer specimen showing high microsatellite instability (MSI-H) and loss of MSH2 protein expression on IHC strongly suggests a germline MSH2 pathogenic variant. Conversely, a microsatellite-stable tumour significantly reduces the likelihood of Lynch syndrome.

Clinical Context

Lynch syndrome detection has important implications beyond the index patient. First-degree relatives of confirmed carriers should be offered cascade genetic testing. Those who test positive benefit from enhanced surveillance protocols, including colonoscopy every one to two years starting at age 20 to 25, and consideration of risk-reducing hysterectomy and bilateral salpingo-oophorectomy for women who have completed childbearing.

Historically, Lynch syndrome was identified using clinical criteria such as the Amsterdam II criteria or the revised Bethesda guidelines. While these criteria remain useful for initial triage, they have limited sensitivity and miss a substantial proportion of carriers. MMRpro provides a more nuanced, quantitative approach that accounts for the full family structure, ages of diagnosis, and tumour marker data, resulting in higher sensitivity for identifying at-risk families.

Understanding PancPRO

Pancreatic cancer carries one of the poorest prognoses of any solid malignancy, with five-year survival rates below 12 percent. Its insidious presentation and rapid progression make early detection and risk stratification especially valuable. While most pancreatic cancers are sporadic, approximately 5 to 10 percent of cases occur in families with a recognisable hereditary pattern.

PancPRO is the BayesMendel model designed to assess familial pancreatic cancer risk. It uses the same Bayesian peeling algorithm as BRCAPRO and MMRpro but models a hypothetical autosomal dominant pancreatic cancer susceptibility gene. The model estimates the probability that an individual carries this susceptibility gene based on their family history of pancreatic cancer, and projects their future cumulative risk of developing the disease.

PancPRO requires the same categories of input as the other BayesMendel models: the pedigree structure, pancreatic cancer diagnoses with ages of onset, current ages of unaffected individuals, and vital status. The model accounts for the fact that pancreatic cancer is relatively rare in the general population (lifetime risk approximately 1.5%), so even a modest family history—such as two first-degree relatives with pancreatic cancer—can substantially elevate the estimated carrier probability.

Individuals identified as high-risk by PancPRO may be candidates for pancreatic cancer surveillance programmes, which typically involve annual endoscopic ultrasound or MRI of the pancreas. Several genes are now known to contribute to familial pancreatic cancer, including BRCA2, PALB2, CDKN2A, ATM, and STK11, and genetic testing panels can complement the PancPRO risk estimate with specific molecular diagnoses.

How Evagene Integrates BayesMendel Risk Models

Running BRCAPRO, MMRpro, or PancPRO has traditionally required manual data extraction from a pedigree, formatting the input for the BayesMendel R package, executing the analysis in R, and interpreting the tabular output. This workflow is time-consuming, error-prone, and creates a barrier between pedigree construction and risk analysis.

Evagene eliminates this friction by integrating all three BayesMendel models directly into the pedigree drawing environment. Family history data entered on the pedigree canvas—cancer diagnoses, ages, relationships, vital status—is the same data the risk models need. There is no separate data entry step.

Technically, Evagene connects to the BayesMendel R package through an R sidecar process—a background R runtime that receives structured pedigree data, executes the requested model functions, and returns results to the web interface. This architecture ensures that the risk calculations are performed by the original, validated BayesMendel code rather than a reimplementation, preserving the statistical rigour and published validation of the models.

From the user's perspective, the workflow is straightforward: draw or import a pedigree, annotate it with clinical data, select which model to run, and view results alongside the pedigree. Carrier probabilities and cumulative risk estimates are displayed in a structured panel that supports clinical interpretation and documentation.

Evagene also supports batch risk screening, enabling clinicians or researchers to run BRCAPRO, MMRpro, or PancPRO across multiple pedigrees in a single operation. This is particularly valuable in research settings or population health programmes where large numbers of families need to be triaged efficiently.

Pedigree data can be imported and exported using GEDCOM format, ensuring interoperability with genealogical databases, other pedigree tools, and institutional health records. This data exchange capability means that family history information collected elsewhere can be brought into Evagene for risk analysis without manual re-entry.

Full documentation on Evagene's risk model integration, including worked examples and parameter reference, is available at the Evagene help centre.

Family-History Scoring Models Alongside BayesMendel

BRCAPRO, MMRpro, and PancPRO answer one specific clinical question: the posterior probability of a pathogenic carrier state given the pedigree. In practice, several other questions arise in the same clinic — what is this woman's lifetime breast-cancer risk from family history alone? Does this family meet a specific testing threshold? Do we suspect Lynch syndrome by classical criteria? Each of these has a dedicated, widely-used scoring model. Evagene implements them all directly, without an R sidecar, on the same pedigree the clinician has already drawn.

Breast-cancer family-history models

• Claus (CASH-derived) — Lifetime and relative breast-cancer risk as a function of affected relatives, degree, and age at diagnosis, with an ovarian-in-first-degree multiplier (Claus, Risch & Thompson 1994). Best for a woman who does not meet high-risk criteria but wants a numerical lifetime estimate.
• Couch — Logistic pre-test BRCA1 probability using average age at diagnosis, ovarian involvement, and Ashkenazi ancestry (Couch 1997), with a 10% testing-threshold flag.
• Frank / Myriad — Empirical BRCA1 and BRCA2 mutation probabilities by canonical family scenario (Frank 2002).
• Manchester Scoring System — Point-based BRCA1 and BRCA2 scores per cancer type and age at diagnosis across the proband and first- and second-degree relatives (Evans 2004). BRCA1 ≥ 16 ≈ 10% carrier probability; combined ≥ 20 ≈ 20%. Used widely by NHS genetics services.
• NICE CG164 / NG101 — Familial breast cancer categorisation into near-population, moderate, or high risk, with a refer-to-genetics flag and the list of triggers matched.
• Gail (NCI BCRAT) — 5-year and lifetime breast-cancer risk using reproductive factors, benign biopsies, atypical hyperplasia, family history, and race/ethnicity baselines (Gail 1989, with updates through 2017).
• Tyrer-Cuzick (IBIS-style approximation) — 10-year and lifetime breast-cancer risk incorporating density, BMI, HRT, LCIS, atypia, and full family history (Tyrer, Duffy & Cuzick 2004). Clearly labelled in-app as a published-algorithm approximation, not the official IBIS binary, whose full coefficients are not publicly released.

Lynch-syndrome family-history criteria

• Amsterdam II — Classical five-point Lynch-syndrome family-history rule (Vasen 1999). Evagene evaluates the pedigree-derivable conditions directly; the exclusion-of-FAP and histology checks are surfaced as clinical-verification notes.
• Revised Bethesda — Triggers for MSI / IHC testing of a colorectal tumour (Umar 2004). Complements Amsterdam II: families that do not meet classical Amsterdam II may still meet Bethesda.

CanRisk / BOADICEA export bridge

The NICE-recommended model for moderate- and high-risk women is BOADICEA, hosted at canrisk.org. It integrates multi-gene panel testing (BRCA1, BRCA2, PALB2, CHEK2, ATM, BARD1, RAD51C, RAD51D, BRIP1), polygenic-risk scores, and mammographic density. Evagene ships a one-click export of a ##CanRisk 2.0 pedigree file populated with pedigree structure, ages at diagnosis, panel test statuses, Ashkenazi flag, and reproductive factors. The clinician uploads the file at canrisk.org and runs BOADICEA there.

BOADICEA is not bundled. The model is licensed by the University of Cambridge; individual clinical use is free after registration, but third-party web-service integration requires a separate commercial licence. The export path is the legally clean route to the gold-standard tool.

Beyond Cancer: Mendelian and Adjacent Risk Models

While cancer risk assessment is a primary use case for pedigree-based risk modelling, the underlying mathematics of Mendelian transmission apply to any heritable condition. Evagene extends beyond the BayesMendel cancer models to provide general-purpose Mendelian inheritance calculators for three inheritance patterns:

• Autosomal dominant — Conditions where a single pathogenic allele is sufficient to cause or predispose to disease. Examples include Huntington disease, Marfan syndrome, and most hereditary cancer syndromes.
• Autosomal recessive — Conditions requiring two pathogenic alleles, such as cystic fibrosis, sickle cell disease, and phenylketonuria. Carrier detection is important for reproductive counselling.
• X-linked recessive — Conditions caused by pathogenic variants on the X chromosome, predominantly affecting males. Examples include haemophilia A, Duchenne muscular dystrophy, and G6PD deficiency.
• X-linked dominant with sex-differential severity — Five sub-modes (equal, males-worse, male-lethal-reproduces, male-lethal-no-reproduction, metabolic-interference males-unaffected) covering Rett-class disorders, incontinentia pigmenti, CFND (EFNB1), and EFMR (PCDH19).
• Mitochondrial (mtDNA) — Strict maternal transmission with sex-differential penetrance and heteroplasmy scaling (LHON, MELAS, MERRF, NARP, Leigh mtDNA subset, Kearns-Sayre, Pearson).
• Digenic two-locus — Classical 25% offspring-affected ratio (Usher type 2, some retinitis pigmentosa, primary congenital glaucoma).
• Imprinting / uniparental disomy — Mechanism-weighted recurrence with the parent-of-origin rule for imprinting-centre defects (Prader-Willi, Angelman, Beckwith-Wiedemann, Silver-Russell, TNDM).
• Polygenic / multifactorial liability-threshold — 20+ common complex diseases (cleft lip ± palate, T2D, schizophrenia, NTDs, asthma …) with empirical Smith / Carter / Harper recurrence tables and Falconer fallback (Carter 1961, Falconer 1965).

These models use the family pedigree to calculate carrier probabilities and recurrence risks, supporting clinical genetics workflows across a broad range of conditions. The same pedigree drawn for cancer risk assessment can be reused for Mendelian, polygenic, or adjacent analysis without additional data entry.

For the full per-model clinical indications, see the decision matrix for clinical geneticists, or the April 2026 release notes for each model's canonical citation.

Frequently Asked Questions

Related Resources