|
Abstract:
|
Simultaneously investigating a multitude of interacting metabolites, proteins, and genes through network analysis gives insights into the development of chronic obstructive pulmonary disease (COPD). Each generated network is then used for downstream analyses to discover association with phenotypes or genetic variants of interest. As networks are multi-dimensional, it is necessary to summarize each network into a quantitative feature, referred to as a network feature. A popular approach is to represent each network with the first principal component of the corresponding abundance matrix. However, this approach does not consider the connectivity between nodes in the network. We present a hybrid approach which incorporates each network’s abundance profile with its topological information. By doing so, the approach has an ability to recover some biological variability among entities that might have been lost during the network construction. The performance of the approach is demonstrated using previously published metabolite-protein networks specific to COPD phenotypes. Additionally, we conduct multiple simulation scenarios to comprehensively evaluate the applicability of the approach.
|
