Phenotypic Substratification of Obesity
Principal Investigator: Lee M. Kaplan, MD, PhD
Director, MGH Weight Center, Massachusetts General Hospital
Current Status: Completed 2010.
Knowledge about the factors that influence body weight regulation is rapidly accumulating, yet our understanding of the etiology of human obesity is still rudimentary. One reason for this lack of clarity is that body weight and composition are affected by a multitude of factors, including physiological, genetic, psychological, environmental and developmental influences. Moreover, obesity is most likely a common symptom of a spectrum of related but pathophysiologically distinct disorders. This heterogeneity is evident in the wide phenotypic variation in parameters such as severity, age of onset, body fat distribution, associated eating behaviors, food preferences, energy expenditure responses, and associated (co-morbid) diseases.
1. What questions did we ask in this DBP?
Using large cohorts of patients with obesity that can be identified from among the large Partners HealthCare System patient data repository, we are asking:
- Can we identify specific clinical phenotypes that, alone or in combination, predict important clinical outcomes and/or response to therapeutic intervention (e.g., medications, surgery, or behavioral modification)?
- Can we identify independent clusters of covariate phenotypic parameters that define clinically meaningful obesity subtypes?
2. How did i2b2 help us answer these questions?
Scientific and technical collaboration with the i2b2 team will allow us to efficiently access, process, and mine data from the electronic medical record. Complementing standard tools of information analysis, the collaborative team will develop, test and apply novel natural language processing (NLP) approaches to examine unstructured text in the medical record. The i2b2 team will also work together to substratify the patient cohorts, using complex queries and statistical approaches (e.g., cluster analysis), to determine clinically meaningful and statistically distinct subtypes.
3. What tools were developed from our work that will be of value to others?
The DBP will require complex and novel query tools to extract relevant obesity-related phenotypes, including co-morbidities of obesity, medications, and continuous clinical traits such as height and weight, from unstructured clinical notes. By using these increasingly sophisticated query tools to extract data from large populations in combination with well-established statistical methods, we hope to significantly accelerate the definition and characterization of clinically relevant obesity subtypes.
4. What new clinical discoveries do we anticipate may arise from our work?
In a disorder as common, heterogeneous and resistant to therapy as obesity, phenotypic substratification can provide several critical advantages. By developing and validating tools that allow real-time refinement of phenotypic characterization and predictive modeling, the DPB can accelerate the development of effective treatment and preventive strategies for this group of disorders. We anticipate that this project will identify clinically relevant obesity subtypes and thereby facilitate the development and characterization of genes and biological pathways that influence body weight and metabolic disease. Through this project, we hope to identify predictors of clinical response to specific therapies, thus helping to accelerate the development and application of new, more effective treatments for obesity and its complications.