Network based stratification (NBS) is a method for stratification (clustering) of patients within a cancer cohort based on genome scale somatic mutation measurements and a gene interaction network. Using a simple random-walk/network-propagation model, NBS transforms a binary vector of tumor’s genomic aberrations (i.e. point mutations/copy-number changes/etc.) to a continuous activation profile on the network. A consensus clustering framework is used to subdivide these activation profiles into distinct subtypes by aggregating multiple clustering results on random subsamples of the data. Clustering can be performed using a non-negative matrix factorization (NMF), network-regularized NMF, or hierarchical clustering.
For more details please see our manuscript:
Hofree, M. et al. Network-based stratification of tumor mutations. Nature Methods doi:10.1038/nmeth.2651 (2013).
The original code accompanying the manuscript can be found here:
nbs_release_v0.2.tgz (V. 0.2 1.4M) .
Additional data and network files are available here:
nbs_v0.2_data.tgz (V. 0.2 4.3G).
A good starting point is looking at the examples provided in the files: demo_NBS_simulated.m or demo_NBS_TCGA.m, included in the release tar-zip linked above.