Medusa is an approach to detect size-k modules of objects (candidate objects) that, taken together, appear most significant to another set of objects (pivot objects).
Medusa operates on large collections of heterogeneous data sets and explicitly distinguishes between diverse data semantics. It builds on collective matrix factorization to derive different semantics, and it formulates the growing of the modules as a submodular optimization program. Medusa is flexible in choosing or combining the semantic meanings, and provides theoretical guarantees about the detection quality.
Large heterogeneous data collections contain interactions between variety of objects, such as genes, chemicals, molecular signatures, diseases, pathways and environmental exposures. Often, any pair of objects---like, a gene and a disease---can be related in different ways, for example, directly via gene-disease associations or indirectly via functional annotations, chemicals and pathways, yielding different semantic meanings.
This repository contains supplementary material for Jumping across biomedical contexts using compressive data fusion by Marinka Zitnik and Blaz Zupan.
The required dependencies to build the software are Numpy >= 1.8, SciPy >= 0.10.
synthetic.py - Demonstrates Medusa on synthetic semantics.
See also scikit-fusion, our module for data fusion using collective latent factor models.
To install in your home directory, use
python setup.py install --user
To install for all users on Unix/Linux
python setup.py build
sudo python setup.py install
To install in development mode
python setup.py develop
@article{Zitnik2016,
title = {Jumping across biomedical contexts using compressive data fusion},
author = {Zitnik, Marinka and Zupan, Blaz},
journal = {Bioinformatics},
volume = {32},
number = {12},
pages = {90-100},
year = {2016},
publisher = {Oxford Journals}
}
Medusa is licensed under the GPLv2.