Improving the Accuracy of Functional Inferences from Links between Proteins Detected by Computational or Experimental Methods

Various methods have been developed for detecting functional couplings among proteins. These include recently developed computational techniques such as the protein phylogenetic profile method [1] and the Rosetta Stone method [2], as well as experimental techniques such as mRNA expression profiling [5]. Here, we present a general approach for quantitatively discriminating more informative functional links from less informative ones. We make the simple assumption that a pair of linked proteins which also share links to other proteins represents a more reliable connection than a pair of linked proteins which share few or no links to other proteins (Figure 1). This approach works on links established by the Rosetta Stone method, the phylogenetic profile method, and on links obtained from mRNA expression profiles.