bihealth / clingen-svi-comp_calibration

Data and code associated with ClinGen-SVI computational sub-group's calibration paper

Code to calibrate tools for clinical interpretation and generate summarized results

This repository contains only the code relevant to the paper. Due to large file sizes, data and intermediate result files are hosted here. Each of the files (data.zip and results.zip) contains a README that provides additional information on the data and results. Please refer to these and link the relevant files to the code provided in this repository so that the scripts and functions access the correct input files.

Repository structure

The repository is organized as follows:

├── LICENSE
├── README.md
├── local_posterior_probability
│   ├── get_all_thresholds.m
│   ├── get_both_bootstrapped_posteriors.m
│   ├── get_both_local_posteriors.m
│   ├── get_discounted_thresholds.m
│   ├── main.m
│   ├── plot_both_posteriors.m
│   └── print_thresholds.m
├── plotting
│   ├── plot_both_posteriors_pub.m
│   ├── plot_correlation.m
│   ├── plot_heatmap_gnomad_set.m
│   ├── plot_heatmap_lr_testset.m
│   └── plot_posterior_wrapper.m
└── results_postprocessing
    ├── assess_default_thresholds.m
    ├── calculate_coverage.m
    └── make_thr_table.m

1. local_posterior_probability

This directory contains the actual implementation of the algorithm to calculate local posterior probabilities (as described in Figure 2 of the paper). The script main.m serves as the wrapper that calls all the other functions in this directory.

Note that print_thresholds.m and plot_both_posteriors.m functions are called in this wrapper mainly to present output immediately for testing and debugging purposes. More advanced, publication-ready versions of these functions can be found in the plotting and results_postprocessing directories.

2. results_postprocessing

This directory contains scripts to post-process outputs from local_posterior_probability and/or generate additional statistics and tables for the results.

• make_thr_table.m : script to generate and systematically print out the score thresholds in Table 2 (and Supplemental Table S1). Note that the format is not exactly as in the paper but it should be easy to update manually to align with the format in the paper.
• assess_default_thresholds.m : script to generate Table 3.
• calculate_coverage.m : script to generate Supplemental Table S2.

3. plotting

This directory contains the code used to make the plots in the paper.

• plot_posterior_wrapper.m : wrapper script to plot Figure 3. This script calls the function plot_both_posteriors_pub.m, which generates each individual local posterior probability plot, i.e., the function is called 26 times for each of the 26 subplots inside Figure 3.
• plot_both_posterior_pub.m : function to plot a single publication-quality local posterior probability plot. Note that this more or less does the same thing that plot_both_posteriors.m in local_posterior_probability does but the resulting plot matches the look and feel of the ones in the paper. It is recommended that this function be used to visualize finalized results.
• plot_heatmap_lr_testset.m : script to plot the heatmap summarizing interval-based likelihood ratios on the validation set (Figure 4A).
• plot_heatmap_gnomad_set.m : script to plot the heatmap summarizing the fraction of gnomAD variants falling within each score interval (Figure 4B).
• plot_correlation.m : script to plot correlation heatmap (Supplemental Figure S1).