search

broadinstitute / PANOPLY

Repository for the Broad Institute Proteogenomic Data Analysis Center (PGDAC) established by the NIH Clinical Proteomics Tumor Analysis Consortium (CPTAC)
Other
33 stars 15 forks source link

Fix NMF clustering in pgdac_main #2

Closed drmani closed 6 years ago

drmani commented 6 years ago

The NMF clustering module adapted from gdac-firecloud eliminates all negative values before clustering. This is not acceptable for log-ratios used in the PGDAC pipeline. Fix or replace module.

karstenkrug commented 6 years ago

The gdac NMF clustering module was replaced by a kmeans-consensus clustering module. Automatic determination of cluster numbers is based on the delta AUC under the consensus CDF as described in the original consensus clustering paper (Monti et al. 2003. Machine Learning).