Skip to content

Missense Score

Summary

The Missense Score (MiS) is a normalized, ensembled metric used to predict the deleteriousness and potential functional impact of missense single nucleotide variants (SNVs). It represents the ratio of in silico prediction algorithms that classify a given variant as deleterious out of 29 established deleterious prediction tools available in the dbNSFP database.

Definition

The Missense Score ($\text{MiS}$) is computed as follows:

$$\text{MiS} = \frac{\text{Number of deleterious predictions}}{\text{Total number of available prediction algorithms}}$$

where the numerator is the count of tools classifying a variant as "deleterious" based on their respective pre-established thresholds, and the denominator is the subset of the 29 algorithms that successfully returned a classification for the variant. The score ranges continuously from $0$ (tolerant across all models) to $1$ (deleterious across all models).

Application and Grouping

Variants are typically grouped into four bins based on their score: - $[0, 0.5]$: Highly tolerant / low deleteriousness - $(0.5, 0.7]$: Moderately tolerant - $(0.7, 0.9]$: Moderately deleterious - $(0.9, 1.0]$: Highly deleterious / high confidence of functional impact

In the large-scale multi-ancestry exome study by Koyama et al. (2026), missense variants in the highest $\text{MiS}$ bins exhibited allele frequency constraint and phenotypic effect sizes (on blood lipid levels) comparable to predicted loss-of-function (pLoF) alleles.

Citations