/kcluster

Machine Learning Toolkit for Multi-omics Cancer Subtype Discovery using Multiple Kernel Learning

Primary LanguagePython

kcluster - Multiple Genomics Cancer Subtype Discovery using Multiple Kernel Learning

kcluster is an open-source toolkit for applying kernel methods to cancer subtype discovery, specifically using Multiple Kernel Learning, k-means clustering, and stochastic optimization to generate subtype clusters for a given cancer dataset.

kcluster provides a Python API that can be run on TCGA (The Cancer Genome Atlas) datasets.

To read more about kcluster, please refer to "Kernel Learning Framework For Cancer Subtype Analysis with Mutli-omics Data Integration" (Bradbury, Lau, Roy 2015).

Dependencies

Python (>=2.6)

scikit-learn (>=0.17)

numpy (>=1.6.1)

scipy (>=0.9)

requests (>=2.8.1)

firebrowse (>=0.1.5)

It is recommended to install these dependencies via the Anaconda package.

If you don't want to use Anaconda, you can install dependencies manually using pip inside a virtualenv