This repository contains the scripts used to generate the findings in the manuscript Joint representation of molecular networks from multiple species improves gene classification. The data for this manuscript were very large and computationally expensive to generate. While the scripts here provide the main code for generating the results, reproducing the data in full requires cluster computating resources and HPC job scheduling code, which are not provided here.
The code was tested with python==3.8.3.
requirements.txt contains packages used for all scripts except those that use pecanpy, whose pacakges can be found in requirements_pecanpy.txt.
The data used in this study is available on Zenodo DOI: 10.5281/zenodo.10246207. To get the data run
sh get_data.sh
The src folder contains the following subfolders with scripts for different stages of the analysis:
-
01: Processing the edgelists into downstream data. -
02: Making the feature matrices used in the machine learning models. -
03: Generating the main findings (Fig 2 and Fig 3). -
04: Generating results from the matched geneset collection (GSC) analyses. -
05: Generating results for training human disease models and looking at predictions across species.
The figure_code folder contains the scripts to generate the figures in the manuscript.