Improving identification of key players in aging via network de-noising

Since human aging is hard to study experimentally due to long lifespan and ethical constraints, current "ground truth" knowledge about human aging has mostly been predicted computationally or statistically by transferring the knowledge from well-studied model species via sequence homology or by studying human gene expression data. Since genes, i.e., their protein products, function by interacting with each other, analysis of protein-protein interaction (PPI) network data in the context of aging is promising. Existing studies of aging from PPI networks have typically relied on static network representations. But because cellular functioning is dynamic, and because different biological data types capture complementary slices of the cell, we recently integrated the static human PPI network with aging-related gene expression data to form dynamic, age-specific networks. Then, we predicted as key players in aging those proteins whose network topologies significantly changed with age. Since current PPI networks are noisy, here, we apply state-of-the-art link prediction methods to the human PPI network to de-noise it. After we show that de-noising improves biological quality of the network, we integrate the de-noised network with the aging-related expression data to construct de-noised age-specific networks and predict novel (hopefully improved) key players in aging from the de-noised data. Indeed, PPI network de-noising improves the quality of the predictions: it results in more significant overlap between the predicted aging-related data and the "ground truth" aging-related data. Yet, we obtain many novel predictions. Thus, we produce new knowledge about human aging from de-noised dynamic networks encompassing multiple biological data types, complementing in this way the existing knowledge obtained from sequence or expression data.