OmniScreen

This repository contains code to reproduce metrics reported in the manuscript and instructions on obtaining raw predictions by the OmniScreen model on MSK and TCGA datasets.

Contents

• metrics.py: Script to compute metrics.
• requirements.txt: Requirements to compute metrics.
• README.md: This file.

Obtaining prediction files

TODO

Reproducing metrics

Requirements

• Python 3.8+
• CPU

Setup

After cloning the repository, install the required packages:

pip install -r requirements.txt

Run the metric computation

To reproduce the metrics computed on MSK data, run:

python metrics.py predictions/msk.csv --type msk --output-file msk_metrics.csv --thresholds-file predictions/thresholds.csv

To reproduce the metrics computed on TCGA data, run:

python metrics.py predictions/tcga.csv --type tcga --output-file tcga_metrics.csv --thresholds-file predictions/thresholds.csv

The metrics files will have the following fields:

tumor_type
tumor_site  # MSK only
label
auc
ap
sensitivity
specificity
ppv
npv
gt_pos_count
gt_neg_count
total_count

TODO

• Link to manuscript
• Link to Zenodo (predictions)