Batch pedigree risk screening: one proband, the whole catalogue, one pass
A clinical overview of batch risk screening — what it is, why it is useful, how it should be used, where it can mislead — and how Evagene implements it across 200+ catalogued conditions with configurable thresholds and integrated Bayesian and Mendelian models.
Short version. Traditional risk modelling is used one condition at a time: a clinician suspects a specific condition, runs the relevant model, and interprets the output. Batch risk screening inverts this: the software sweeps the entire catalogued disease list for the proband, computes the appropriate model for each, applies configurable thresholds, and surfaces only the conditions where the family history crosses a meaningful threshold. It is a discovery and triage tool — a way to reduce the chance that a relevant condition is missed because it did not happen to come to mind — rather than a replacement for focused assessment. Evagene implements batch screening across its 200+ catalogue with BayesMendel and Mendelian models, configurable thresholds, and programmatic access via REST API and MCP.
What batch risk screening is and is not
"Batch" here does not mean "multiple probands at once" — though that is a separate use case. It means "all catalogued conditions at once for a single proband." A single pedigree, with its annotated diagnoses, ethnicity, consanguinity, and family structure, goes through the full set of risk and inheritance analyses in a single run. Each condition is scored with its own model and compared against a threshold. Conditions below threshold are silent; conditions above threshold surface as flags the clinician reviews.
It is not a new statistical method. The risk models themselves — BRCAPRO, MMRpro, PancPRO, Mendelian inheritance calculators — are the same as those used in focused assessment. What is new is the orchestration: applying the full catalogue to the pedigree in one pass and presenting a triaged output, rather than asking the clinician to choose a model per condition.
It is also not an unsupervised diagnostic system. Every flag is reviewed by a clinician, who decides whether to act on it (deeper modelling, phenotyping, testing, referral) or to dismiss it in context.
Why batch screening matters clinically
Three reasons account for most of the clinical value.
Catalogue breadth exceeds clinician memory. A well-prepared clinical geneticist holds a mental catalogue of perhaps a few hundred Mendelian conditions, with the most common or the most recently read occupying the most space. The full landscape of Mendelian disease is larger than any single clinician can actively consider during any single consultation. Batch screening widens the aperture without requiring recall.
Family-history signals cluster in non-obvious ways. A pedigree may have features that, considered singly, do not point anywhere, but that combined suggest a specific condition. A grandmother with colorectal cancer at 52, an aunt with endometrial cancer at 45, and a second cousin with an early colorectal tumour might not individually prompt Lynch syndrome thinking; together they often should. Batch screening applies the combination test rather than the individual test.
Standardisation across clinicians. Different clinicians working in the same service have different mental catalogues. A screen run by one clinician produces the same flags as a screen run by another, so the minimum floor of consideration is consistent across the team.
None of this replaces focused expertise. When a clinician knows what to suspect, they should run the appropriate model directly. Batch screening is a safety net, not a primary workflow.
Models used in a batch screen
Evagene's batch screen combines validated Bayesian risk models for cancer with Mendelian inheritance calculators for non-cancer monogenic conditions.
- BRCAPRO. Bayesian model for BRCA1 and BRCA2 variant probability, based on breast, ovarian, pancreatic, and male breast cancer in the pedigree. See our hereditary cancer risk assessment guide for the underlying method.
- MMRpro. Bayesian model for MLH1, MSH2, and MSH6 variant probability in Lynch syndrome, based on colorectal, endometrial, and Lynch-spectrum cancers in the family.
- PancPRO. Bayesian model for familial pancreatic cancer.
- Mendelian inheritance calculator. Autosomal dominant, autosomal recessive, and X-linked recessive carrier and affection probabilities for conditions in the 200+ catalogue. See the Mendelian inheritance calculator guide for details.
Each condition is scored with the most appropriate model for its inheritance class. The output for each condition includes the probability, the threshold used, and the pedigree evidence that drove the calculation.
Thresholds: the art of calibration
The choice of threshold is the pivot of the screen. Too permissive and the screen produces a long list of low-signal flags that the clinician has to triage anyway; too strict and the screen risks missing conditions that warranted a second look. A few practical principles:
- Different conditions warrant different thresholds. A 10% BRCA carrier probability might flag for genetic testing; a 10% probability for a very rare Mendelian condition might reflect a thin family history rather than genuine elevated risk. Thresholds should be set per-condition, not globally.
- Testing-eligibility criteria are natural thresholds. Many services use published eligibility criteria (NICE guidelines, NCCN, local policy) as the threshold for surfacing a flag. Evagene allows thresholds aligned to testing-eligibility criteria where those are established.
- Small families need higher thresholds. A pedigree with few informative relatives is noisier; models are more prior-driven and more prone to false positives. Raising the threshold for small families reduces the noise.
- Thresholds should be reviewed periodically. A service that runs batch screening routinely should revisit its thresholds as new evidence and new testing guidelines are published.
Evagene supports configurable thresholds per model and per condition, with reasonable defaults that can be tuned for a service's local context.
A disciplined workflow
The value of batch screening is realised only when its outputs are reviewed with discipline. A good workflow looks like this.
- Construct the pedigree carefully. Batch screening amplifies whatever is in the pedigree, so errors or omissions are amplified too. A thin or inaccurate pedigree produces misleading output, whether screened or not.
- Set thresholds in advance. Decide the service's thresholds before running the screen, not after seeing the output. Thresholds chosen in response to output drift toward confirming what the clinician already expected.
- Run the screen. In Evagene, this is a single action; the screen runs across the catalogue and returns a flagged list.
- Review the flagged list. For each flag, examine the evidence: which relatives drove the probability, whether ages of onset are consistent, whether the ethnicity and consanguinity inputs are correct. Dismiss flags that are artefacts (e.g., a small family where the Bayesian prior dominates).
- Deep-dive where warranted. For flags that survive review, run the focused model (BRCAPRO, MMRpro, PancPRO) on its own, discuss with the patient, and consider testing or referral.
- Record the screen outcome. Keep a record of the screen run, the thresholds used, and the decisions taken. This protects the service if the screen is revisited later, and supports quality improvement.
Pitfalls to watch for
Prior-probability calibration. For very rare conditions, Bayesian models use population prior probabilities that may not match the population in front of the clinician. A flag in a specific ethnic group should be interpreted with that group's known prevalence in mind.
False positives on small families. A pedigree of seven individuals produces noisier output than one of seventy. Small families are more prone to both false positives (flags driven by the prior rather than the data) and false negatives (real risk masked by limited information).
Over-reliance on the screen. Batch screening does not replace clinical history-taking, phenotyping, and judgement. A pedigree that misses a key second-degree relative will produce misleading output however sophisticated the screen. The clinician's scrutiny of the pedigree itself matters more than the screen.
Threshold gaming. Tightening thresholds to reduce the review workload can mask genuine signals; loosening them to catch more can overwhelm the clinician. Choose thresholds deliberately and revisit them with evidence.
Polygenic and complex conditions. Many common diseases have substantial polygenic contributions that Bayesian Mendelian models do not capture. Batch screening across a Mendelian catalogue is not a substitute for polygenic risk assessment where that is relevant.
How Evagene implements batch screening
Evagene's batch risk screening sweeps the 200+ disease catalogue against the active pedigree in a single run. For each condition it uses the appropriate model — BRCAPRO, MMRpro, or PancPRO for cancer conditions within their scope; Mendelian inheritance calculation for autosomal dominant, autosomal recessive, and X-linked recessive conditions; and Bayesian priors for conditions in the catalogue. Configurable thresholds per model and per condition let a service calibrate to local testing-eligibility criteria and population context.
The output is a flagged list, with each flag showing the calculated probability, the threshold used, and the specific pedigree evidence that drove the flag (which relatives, which diagnoses, which ages of onset). Clicking into a flag produces a deeper view of the model output, including confidence intervals where relevant. The clinician can mark each flag as actioned, dismissed, or deferred, producing an audit trail of the screen's contribution to the case.
Batch screening is available through the Evagene interface, through the REST API, and through the MCP server, so it can be embedded in a clinical agent workflow or an automated review process. AI interpretation (via bring-your-own-key LLMs) can be applied to the screen output to draft a narrative summary for inclusion in a clinic letter. As with every Evagene feature, the screen output is a drafting aid, not a clinical decision.
Frequently asked questions
What is batch pedigree risk screening?
Running a risk or inheritance analysis across the full disease catalogue for a single proband in one pass, with configurable thresholds to surface only conditions where the family history meets a meaningful level.
Why is it clinically useful?
It widens the aperture beyond the clinician's active memory, catches conditions where combined signals in the pedigree would not trigger any single one on its own, and standardises the minimum floor of consideration across a team.
What models does the screen use?
BRCAPRO, MMRpro, and PancPRO for cancer; Mendelian inheritance calculators for AD, AR, and XR conditions; applied across the 200+ catalogued diseases using the appropriate model per condition.
What are the pitfalls?
Prior-probability calibration for rare conditions; false positives on small families; over-reliance on the screen at the expense of clinical history-taking; threshold gaming; limited applicability to polygenic conditions.
When should I not use batch screening?
When a focused assessment is the right tool for the specific question, or when the pedigree is too thin to produce meaningful signal. Batch screening is a discovery and triage tool, not a primary workflow.
How does Evagene implement it?
Configurable thresholds per model and per condition, a single-run sweep across the 200+ catalogue, a flagged output with the evidence that drove each flag, and programmatic access via REST API and MCP for AI-agent orchestration.