Skip to content

DANN score

DANN, short for Deleterious Annotation of genetic variants using Neural Networks, is a genome-wide model for single-nucleotide variants. It combines hundreds of genomic annotations to compare observed human variants with simulated variants used as a proxy for potentially deleterious changes. A higher score is a stronger model signal. It is not a diagnosis or a calibrated probability that a person has, will develop, or pass on a condition.

Score range in Gene Inspector Pro

DANN scores run from 0 to 1. The original method does not set a clinical cutoff. Gene Inspector Pro uses these display bands for review priority:

Score Display meaning
Below 0.5 Lower DANN signal
0.5 to below 0.96 Intermediate DANN signal
0.96 or higher Higher DANN signal

The 0.96 boundary is a Gene Inspector Pro review convention. It is not a published diagnostic threshold and should not be read as a benign or pathogenic classification.

What the score can cover

DANN can score coding and non-coding single-nucleotide variants. That wider coverage does not make it disease-specific. Its training labels are proxies: some observed variants can have effects, and some simulated variants can be harmless.

How to use it

Use DANN to prioritize a variant for closer review. Check the result with variant consequence, population frequency, ClinVar submissions, inheritance, genotype quality, functional evidence, and the clinical question. A higher DANN score does not establish that a variant causes disease.

Sources