Clustering and Sorting: Applications to Biological data

Eytan Domany
Dept of Physics of Complex Systems
Weizmann Institute of Science
Rehovot, Israel

I describe briefly three methods for unsupervised data analysis: 
SuperParamagnetic Clustering (SPC), Coupled Two Way Clustering (CTWC) 
and Sorting Points Into Neighborhoods (SPIN). I demonstrate how these 
methods were aplied to analyze expression data from cervical cancer 
and colon cancer. In the first case a "proliferation cluster" of 160 
genes were discovered, whose expression level was indicative of outcome. 
For colon cancer several gene clusters that correspond to different 
functional categories were identified, and clean metastasis samples 
were separated from those that were contaminated by surrounding host tissue.