November 7, 2014
To add to our understanding of the causative aspects of rare disease, a group from Case Western Reserve University has published the creation of a disease phenotype network that aids in linking complex clinical phenotypes. This article, published in Journal of Biomedical Informatics, describes a novel method of creating phenotype network database which does not rely on mining textual phenotype descriptions, but on the usage of highly accurate disease-manifestation semantic relationships from Unified Medical Language System (UMLS). Chen et al., have called this database Disease Manifestation Network (DMN) and made it available publicly at nlp/case.edu/public/data/DMN.
According to the authors the usage of 50 543 highly accurate disease-manifestation semantic relationships UMLS helped capture major aspects of disease phenotypes which can successfully predict disease causes. A salient feature of this phenotype network database included that DMN not only contained existing knowledge but also some novel insights which the authors found by comparing DMN and mimMiner (a phenotype network database constructed through text mining). The authors also found that DMN partially correlated with the genetic network database – Human Disease Network (HDN) – based on Online Mendelian Inheritance in Man (OMIM) and Genome-Wide Association Studies (GWAS). Finally, using the example of Marfan Syndrome the authors found that DMN has the potential to provide“new leads to discover unknown causes of Marfan Syndrome”, thus concluding that a combinatorial approach where mimMiner and DMN disease is used would be an excellent method for gene discovery and drug.