Abstract

 

Compared to the initial expectation human beings are gene-poor organisms. Many genes and pathways are likely to play a role in more than one disease, and numerous examples of gene pleiotropy and protein multi-functionality presumably await discovery.  This situation contributes to the recent interest in clinical healthcare sector data and their accounts of fine-grained multi-morbidities. Patient record data is a potentially rich data source for discovering correlations between diseases, drugs and genetic information in individual patients. A fundamental question in establishing biomarker-phenotype relationships is the basic definition of phenotypic categories. As an alternative to the conventional case-control, single disease model the talk will describe attempts to create phenotypic categories and patient stratification based on longitudinal data covering long periods of time. We carry out temporal analysis of clinical data in a more life-course oriented fashion. We use data covering 6-7 million patients from Denmark collected over a 20 year period and use them to “condense” millions of individual trajectories into a smaller set of recurrent ones. This set of trajectories can be interpreted as re-defined phenotypes representing a temporal diseaseome as opposed to a static one computed from non-directional comorbidities only. We present examples, including one from the area of inflammatory disease area where five diseases seem to co-occur primarily due to shared loci rather than follow-on disease etiologies.  This type of work can potentially gain importance in projects involving population-wide genome sequencing in the future.