Article

AlphaMissense and DANN scores

AlphaMissense estimates missense variant impact from protein structure and conservation. DANN assigns 0 to 1 scores across genomic features.

Protein structure evidence view connected to abstract genomic score tracks.

Gene Inspector Pro shows AlphaMissense and DANN scores. AlphaMissense estimates whether a missense variant is likely to damage protein function. DANN assigns each variant a score from 0 to 1 based on genomic features.

What is AlphaMissense?

AlphaMissense is an AI model from DeepMind that estimates the pathogenicity of missense variants. It uses protein structure and evolutionary conservation instead of training directly on human clinical database labels. Read the original published paper on AlphaMissense.

Illustration of AlphaMissense using protein structure and evolutionary conservation to estimate variant impact — AlphaMissense scores a missense change using protein structure and evolutionary conservation.

How it works

AlphaMissense evaluates the changed amino acid, the protein position, and conservation across species. This helps it score rare or novel missense variants that may be absent from curated databases.

Comparison of database-trained variant prediction and AlphaMissense structure plus evolution evidence — Database-trained predictors depend on prior labels and annotations. AlphaMissense uses structural and evolutionary features.

In Gene Inspector Pro

You can find AlphaMissense scores under the AMS label in the variant details. The High AlphaMissense Score panel gathers variants with score of at least 0.5.

What is DANN?

DANN (Deleterious Annotation of genetic variants using Neural Networks) predicts whether a variant is likely to be disease-causing. The method uses deep neural networks and genomic features, and it can score coding and non-coding variants. Read the published DANN paper.

How it works

DANN scores run from 0 to 1. Higher values mean a higher predicted chance of a harmful effect. The Gene Inspector help guide uses a score higher than 0.96 as a high-score review threshold.

In Gene Inspector Pro

Gene Inspector Pro shows DANN next to other predictor values and includes a high-score review panel. Treat it as one prioritization signal, then check consequence, allele frequency, ClinVar, inheritance, genotype quality, and source evidence.