Helix Insight

Documentation / Literature Evidence / Evidence Strength

Evidence Strength

Each publication in the literature results is assigned an evidence strength category based on the type of evidence it provides. These categories are aligned with ACMG criteria to help the geneticist quickly identify which publications are most relevant for classification decisions.

Strength Categories

StrongPS3 / PP3 candidate evidence

The publication describes the exact variant found in the patient and includes functional evidence (animal model, cell assay, enzyme activity). This is the highest-value literature finding for ACMG classification.

Criteria: Exact variant match AND functional study data present

ModeratePM-level supporting evidence

The publication either describes the exact variant (without functional data) or provides functional evidence for the gene (without the exact variant). Either finding contributes meaningful evidence to classification.

Criteria: Exact variant match OR functional study data present

SupportingPP4 / PP5 candidate evidence

The publication discusses the same gene in the context of phenotypes that overlap with the patient’s presentation. Supports the genotype-phenotype correlation without variant-specific or functional evidence.

Criteria: Gene match AND phenotype match present

WeakBackground reference

The publication mentions the gene but lacks phenotype overlap, variant specificity, or functional data. Useful as background context but insufficient for direct ACMG evidence application.

Criteria: Gene mentioned only

How Functional Data Is Detected

The platform identifies functional study evidence through two methods:

MeSH Descriptor Matching

Publications indexed with MeSH terms indicating model organisms (zebrafish, mice), knockout studies, cell line experiments, or molecular biology techniques are flagged as containing functional data.

Abstract Keyword Fallback

When MeSH descriptors are insufficient, the abstract text is searched for functional study keywords. This catches publications where the functional component is described in the text but not captured in MeSH indexing.

How Variant Matching Works

Variant mentions extracted during database ingestion are compared against the patient's variants. An exact match requires the same gene symbol and the same HGVS notation (cDNA or protein level). The platform recognizes multiple notation formats:

HGVS cDNA: c.123A>G, c.456_789del, c.100_101insATG

HGVS protein: p.Arg248Gln, p.Gly12Val

Legacy single-letter: R248Q, G12V

When the exact variant is not found but the gene is mentioned, the publication receives a partial variant score (0.3 instead of 1.0) and is categorized based on its other evidence components.

Clinical Judgment Required

Evidence strength categories are computational estimates based on extracted data. A publication labeled "Strong" may contain a variant match in a different clinical context than the current patient. A "Weak" publication may contain a crucial observation in its discussion section that the extraction pipeline did not capture. The geneticist should review the actual publication before applying evidence to ACMG classification.