B0479
Title: Differential networks: Interpretation of an edge
Authors: David Macleod - London School of Hygiene and Tropical Medicine (United Kingdom) [presenting]
Abstract: A differential network refers to an analysis method which compares two networks each representing one of two subgroups. For example we can compare the metabolomic network formed by a set of diseased individuals with the network formed by a set of disease free individuals, with an aim of identifying structural differences which could provide information about the aetiology of disease. The primary method of estimating this network is to estimate the difference in partial correlations of each pair of nodes in each of the two subgroups, with a suitably large difference leading to the inclusion of an edge in the differential network. Most of the recent literature concerning this method focuses on the method of estimating the edges in the network. However, little time has been spent on interpreting what an edge in a network of this type means. A simulation study has been performed, investigating how the method responds to a particular type of data where a pair of nodes in the network are causes of disease to begin to develop our understanding of what a differential network represents.
