Analysis of incomplete gene expression dataset through protein-protein interaction information.

This paper shows a graph based method to analyze proteomic expression data. The method allows the prediction of the expression of genes not measured by the gene expression technology based on the local connectivity properties of the measured differentially expressed gene set. The prediction of the expression jointly with the stability of this prediction as a function of the variation of the initial expressed set is computed. The method is able to correctly predict one third of the proteins with independence of variations on the selection of the initial set. The algorithm is validated through a Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometer (MALDI-TOF) protein expression experiment aiming the study of the protein expression patterns and post-translational modifications in human endothelial vascular cells exposed to atherosclerotic levels of Low Density Lipoproteins (LDL).