acmgscaler

The goal of the acmgscaler R package is to provide a robust approach for gene-level calibration of variant effect scores, such as computational predictions or functional assay results, against ACMG/AMP evidence thresholds. The package is lightweight and written entirely in base R, without additional dependencies.

Colab notebook

A plug-and-play Google Colab notebook with a simple interface is available for all users.

Installation

You can install the stable version of acmgscaler directly from GitHub using the devtools package:

# install devtools if you haven't already
install.packages('devtools')

# install the acmgscaler package from GitHub
devtools::install_github('badonyi/acmgscaler')

Quick start

library(acmgscaler)
data('variant_data', package = 'acmgscaler')

# calibrate example data
calib <- calibrate(
  df = variant_data,
  value = 'score', 
  prior = 0.1,
  group = 'gene'
)

# likelihood_ratio data for each variant
calib$BRCA1$likelihood_ratios

# score thresholds for ACMG/AMP evidence levels
calib$BRCA1$score_thresholds

The Colab uses a non-exported internal function to display the score intervals:

acmgscaler:::prettify_score_thresholds(calib$BRCA1$score_thresholds)

  ACMG/AMP evidence strength      score
1          Benign-VeryStrong       <NA>
2              Benign-Strong > -0.61243
3            Benign-Moderate > -0.77564
4          Benign-Supporting > -0.84782
5      Pathogenic-Supporting < -1.01955
6        Pathogenic-Moderate < -1.08209
7          Pathogenic-Strong < -1.17951
8      Pathogenic-VeryStrong < -2.14538

How to cite acmgscaler

If you find this package useful, refer to Badonyi & Marsh, bioRxiv (2025); doi: 10.1101/2025.05.16.654507v2