The Yeast Interactome -- analysis and evaluation of diverse sources of information.

Jeremiah J Faith1, Ravi Sachidanandam2
1faith@cshl.org, Cold Spring Harbor Laboratory; 2sachidan@cshl.org, Cold Spring Harbor Laboratory

Protein-protein interactions in yeast have been studied using diverse techniques, such as yeast two-hybrid, mass spectrometry, and gene knockouts. Combining the data from these sources has been problematic, since their effects on the network topology have not been fully characterized, nor has their reliability been firmly established. We have developed tools, based on monte-carlo techniques and spectral methods from graph theory, to evaluate the relative importance of various interaction detection methods in characterizing and predicting functional groups of genes. We present a comparison of mass-spec versus yeast two-hybrid data for several functional protein families as well as groups of proteins involved in a well-studied pathways, such as glycolysis. We also present inferences on the reliability of different datasets based on our analysis as well as the pros and cons of mixing datasets derived from radically different experiments.