Tutorial

Confidence needs evidence.

QuantBayes Studio takes a result from a pipeline, human review, or AI assistant and adds quantified evidence sufficiency under a declared evidence profile.

The result remains the same object. QuantBayes adds the evidence layer: what was checked, what was missing, and how much support was present under the selected profile.

Select an example profile

Start with a community profile, or build a private one

Evidence profiles define what must be checked. Communities can publish shared profiles for each domain, while teams can keep private profiles for internal standards.

Selected: Clinical genetics

Active example

Rare disease trio WGS verification

A genetics pipeline returns candidate causal variants. QuantBayes Studio adds evidence sufficiency, showing which candidate has the most complete, inspectable evidence trail under a community clinical genetics evidence profile.

Clinical genetics researcher or doctor reviewing sequencing data, reports, and laboratory material in a standard genetics analysis setting.
User context
High-level clinical genetics workflow showing trio sequencing input, candidate variants, evidence checks, and a ranked evidence result.
Evidence workflow
Before
DNA analysis result
Unquantified
RYR2 frameshift, variant_002_RYR2

A rare disease pipeline has nominated RYR2 frameshift as a candidate. The result is plausible, but the evidence support has not yet been quantified.

After QuantBayes
DNA analysis result
Evidence quantified
RYR2 frameshift, variant_002_RYR2
4/5 evidence checks complete
θ 0.714, 95% CrI 0.359 to 0.957

The same candidate result is retained. QuantBayes adds evidence sufficiency, uncertainty, and a reusable evidence trail under the selected evidence profile.

Top result

RYR2 frameshift

0.714
Evidence confidence
0.359 to 0.957
95% CrI
4 of 5 checks complete

Most complete evidence trail in this run. This does not prove causality.

Ranked result

Evidence support by item

RankItemEvidenceMissingResultEvidence confidence (95% CrI)
1
RYR2 frameshift
variant_002_RYR2
4/51Most complete0.714 (0.359-0.957)
2
CFTR p.Phe508del
variant_001_CFTR
2/53Partial evidence0.429 (0.118-0.777)
3
MYH7 missense
variant_003_MYH7
2/53Partial evidence0.429 (0.118-0.777)
4
TTN intronic
variant_004_TTN
2/53Partial evidence0.429 (0.118-0.777)
5
NOD2 common risk allele
variant_005_NOD2
1/54Weak evidence0.286 (0.043-0.641)

Evidence profile

The profile is selected before scoring

The evidence profile defines the checks that will be applied. Evidence is added later by a user, pipeline, or AI assistant.

Evidence checkWhat it asks
Inheritance compatibilityDo family genotypes fit the proposed inheritance pattern?
Population rarityIs the variant rare enough for a rare disease model?
Disease gene relevanceIs the gene relevant to the clinical presentation?
Functional gene contextIs there biological support for the gene or region?
ClinVar assertion completenessIs the variant directly supported by a strong public assertion?

Evidence trail

Evidence trail for RYR2 frameshift

The score is useful because the underlying evidence remains inspectable.

EvidenceWhat was foundSourceVerifiable evidence
InheritanceProband is heterozygous and both parents are reference.Trio genotype comparisonPresent
Population frequencyVariant is absent or very rare in population reference data.Population annotationPresent
Gene contextGene is biologically plausible for a cardiac phenotype.Disease-gene profilePresent
Protein contextGene encodes a large functional calcium channel protein.Protein annotationPresent
Public assertionDirect public assertion is limited or uncertain.ClinVar-style assertion fieldMissing

Scale

Useful for one result or one thousand agents

A single result can be audited. A thousand agents can be compared by the same evidence profile, so the safest supported result can be returned.

1
Result

Check one manual, pipeline, or AI-generated result.

1,000
Agents

Score many agents against the same evidence profile.

1
Return

Select the result with the best evidence trail.

Behind the scenes

Only two things need the user’s judgement

The user or organisation chooses the evidence profile and supplies or reviews the evidence. Studio records and calculates the rest.

TypeObjectWhat it does
User inputEvidence profileThe independent profile selected by the user or organisation.
User inputEvidenceWhat the user, pipeline, or AI assistant found for each item.
System recordProjectGroups related work.
System recordCaseHolds the candidates, evaluations, strategies, targets, or items.
AutomatedRunFreezes the profile, evidence, matrix, parameters, and engine version.
AutomatedResultStores the ranking, uncertainty, evidence trail, and report.

Start

Turn a result into an auditable evidence object.