This paper in Nature from Ian Taylor and Jeff Wrana from U. Toronto present a compelling method for the discrimination and diagnosis of disease states based on the disruption of co-expression patterns with neighbors in a protein interaction network. Typically, the search for cancer biomarkers involves searching for differentially expressed genes which alone can discriminate between different cancer diagnoses. The authors present a method which integrates co-expression and public protein interaction information to show that genes whose protein network neighbors become dysregulated may be used to predict breast cancer outcomes. Their findings support the emerging notion that biomarkers might be improved by focusing on pathways rather than single genes alone. A couple of recent papers from our lab indicate show early signs of promise in this regard. Their results have implications for the systematic discovery of markers in a variety of diseases.