Patrícia C. T. Gonçalves1, Pedro Campos1,2
Many diseases are related to several aspects, such as heredity, geography, demography, social aspects, etc. It is possible to visualize relationships in the form of social networks, in which the nodes of the network represent patients, and the edges of the network are family ties between those patients, for example.
Social networks enable the analysis of complex data of this type, and the level of complexity can increase even more when several variables are included simultaneously, such as geographic location, kinship, age or type of disease. Here, the use of multilayer networks can be an advantage.
A multilayer network is a representation of a connected system of agents that may interact in different modes. Those different modes, or variables, are represented by layers.
There are several types of patterns that can be extracted from social networks. Network centrality, for example, measures the level of importance of the nodes in the network. Community measures evaluate the network partition into modules, thus identifying individuals with high degrees of homogeneity.
Our main goal is to relate patient data through networks of individuals to detect possible “hidden” links between variables. More precisely, we want to relate cancer patients through their kinship degree, age, type of tumour and geographical location. Using centrality and community measures we hope to identify patterns within those data.
Thus far, we have been using fictitious data to try out our approach. We are involved in contacts with IPO to have access to real data, or resort to data simulation, based on empirical evidence.
keywords: Data Analysis, Health Data, Multilayer Networks, Pattern Identification