Overview

This repository accompanies the SCoNE repository (https://github.com/G2Lab/SCoNE) from the paper 'Sparse Covariate-aware Non-negative Extraction Improves Complex Disease Subtyping', and was used to run SCoNE and other unsupervised phenotyping methods on the All of Us Researcher Workbench. This code can be run by authorized users on the Researcher Workbench.

Directory Structure

algorithms/    Unsupervised phenotyping algorithms
evaluation/    Evaluate reconstruction error
01*/     Setup resources to generate clinical, genetic and covariate matrices for any task
02*/     Validate SCoNE vs. other methods for the task of predicting hypertension (case-control decomposition)
03*/     Explore SCoNE findings on major depressive disorder (case only decomposition)
*_utilities.py Various functions to run unsupervised phenotyping methods and evaluate the resulting decompositions

Notes

02b-Decomposition.ipynb is used to run decomposition for both case-control and case-only.