The largest publicly available ‘network database’ of observable traits for more than 10,000 diseases was published online this month in PLoS Computational Biology. Looking at links between co-existing or related diseases, together with genotypic and proteomic data, can offer a new way of understanding how diseases develop and progress, according to César Hidalgo and colleagues.

A single disease can conceal a host of complex biological processes and a range of causes. This gap can be bridged by capturing connections between the artificially discrete disease categories, the authors explain.

“Exploring comorbidities from a network perspective could help determine whether differences in the comorbidity patterns expressed in different populations indicate differences in biological processes, environmental factors, or health care quality provided for each population,” writes Hidalgo, from Harvard University in Massachusetts, USA, and colleagues. “The PDN [Phenotypic Disease Network] could be the starting point of studies exploring these and related questions.”

Built with information gleaned from the disease history of more than 13 million elderly patients in the USA, the database contains four years’ worth of data on all diagnoses recorded on more than 30 million medical insurance claims in standard ICD9-CM (International Classification of Diseases) code. The data were grouped by the patients’ race and gender.

The authors’ first analysis revealed that patients tend to develop diseases connected to conditions they already have. Those patients who suffer from diseases with a large set of network connections tend to die sooner than those in a smaller ‘local’ network of related conditions, according to the analysis. But this progression of disease varies depending on gender and racial background.

Joel Dudley, Bioinformatics Programmer at the Stanford Center for Biomedical Informatics Research in the USA, is cautious about the value of the database for identifying and preventing disease. It is an excellent and novel resource, made freely available by the researchers, he says, “but the direct impact to clinical research and practice is likely to be limited”.

The PCN offers a unique perspective and opens the door to new directions in the application of health research, says Dudley, but the advantage it offers to clinical practice needs further evaluation. “It is difficult to ascertain the value of the relationships presented in this network without some formal, quantitative comparison to current clinical understanding,” Dudley explains. “The clinical literature is rich with co-morbidity research, and therefore [a comparison] is possible.”

As it stands, the method shows promise for identifying disparities between populations, suggests Dudley. “Also, the idea of analysing disease propagation along the comorbidity network is intriguing.” At the same time the database is limited in some ways, he notes. It’s focussed on patients over 65 years of age, relies on a potentially biased disease-coding system, and lacks information on drug side-effects or other easily identifiable factors that could be driving the disease inter-relationships.

“The next step in this line of work is to integrate these findings with more clinically relevant data... to gain a more complete systems understanding of the factors leading to disease,” says Dudley.

Network approaches to the study of human disease using data of this type have been published before, according to Dudley. The authors say that large network databases are available for disease-associated genes and proteins. But until now, difficulties with access to extensive medical records have prevented researchers from doing the same for a large volume of diseases diagnosed in the population.

This database is “by far” the largest of its kind and carries a wealth of clinical information, comments Charles Auffray, Research Director of the Centre National de la Recherche Scientifique in Villejuif, France. “[The authors] provide a dynamic overview of the landscape of these human diseases, which were often considered individually or in closely related groups.” Analysis of these data can help advance treatment, he points out, by targeting a number of different pathways involved in the development of disease.