Phenotypic disease network (PDN)
The first phenotypic disease network was constructed by Hidalgo et al. to help understand the origins of many diseases and the links between them. Hidalgo et al. defined diseases as specific sets of phenotypes that affect one or several physiological systems, and compiled data on pairwise comorbidity correlations for more than 10,000 diseases reconstructed from over 30 million medical records. Hidalgo et al. presented their data in the form of a network with diseases as the nodes and comorbidity correlations as the links. Intuitively, the phenotypic disease network can be seen as a map of the phenotypic space whose structure can contribute to the understanding of disease progression.
History
During the last decade, several papers were published that aim at understanding the origins and interrelatedness of diseases using the analytical tools of network science. Interactions between disease-associated genes, proteins, and gene expressions have been explored.However, phenotypic information was essentially overlooked, despite the fact that there exist extensive, high-quality data on it in the form of clinical histories, until the seminal paper of Hidalgo et al. introducing the human phenotypic disease network.
Data and methodology
Comorbidity correlation measures
Comorbidity measures are used to measure the "distance" between two diseases. The relative risk of observing disease i and j affecting the same patient is given bywhere denotes the number of patients affected by both diseases, N is the total number of patients in the population, and and are the prevalences of diseases i and j, respectively.
The -correlation can be expressed as
Both measures have inherent biases: RR overestimates relationships involving rare diseases and underestimates the comorbidity between highly prevalent diseases, while the -correlation is accurate in describing comorbidity between diseases with similar prevalence but underestimates the comorbidity between rare and common diseases. In the PDN, nodes are disease phenotypes and links connect those phenotypes that have significant comorbidity correlation according to the RR and -correlation.
Considering the complementary biases of these two measures, Hidalgo et al. constructed a separate PDN for each.
Disease network dynamics
Disease progression
Comparing the average correlation between illnesses diagnosed in the first two visits and those diagnosed later to the average correlation in a randomized control case, inter-visit correlations were found to be significantly larger than those that would occur by chance alone, pointing to the fact that patients develop diseases that are close in the PDN to those they already have.Connectedness and mortality
Hidalgo et al. also established a connection between that the mortality associated with a given disease and its connectivity in the PDN. Diseases that are preceded by others are usually more connected than those that precede others, and they tend to be more lethal.That is to say, patients that have a disease that is more connected in the network face higher mortality rates that those patients who have less connected conditions.