David M Maslove1, John C Marshall. 1. *Departments of Medicine and Critical Care Medicine, Queen's University, Kingston †Critical Illness and Injury Research Centre, Li Ka Shing Knowledge Institute, St. Michael's Hospital ‡Department of Surgery, University of Toronto, Toronto, Ontario, Canada.
Abstract
RATIONALE: Most gene expression studies of sepsis have used either whole blood or specific leukocyte fractions as source tissues for RNA. Data regarding the relative utility of these different tissue sources are lacking. OBJECTIVES: To evaluate the utility of different source tissues in studying gene expression in sepsis. METHODS: We undertook a systematic analysis of sepsis gene expression studies, including both adult and pediatric cohorts. We used clustering methods to partition samples according to gene expression levels, and compared expression cluster labels to clinical diagnoses. We also quantified the strength of cluster formation based on expression data from different tissue sources using average silhouette widths as a measure of cluster cohesiveness. RESULTS: We included 22 separate expression datasets. Whole blood was used as the source tissue in 15 studies, while leukocyte isolates were used in seven studies. Whole blood samples yielded greater specificity for the diagnosis of sepsis than data from leukocyte isolates (94% vs 78%, P = 0.03). Whole blood-derived data also yielded more cohesive clusters (median silhouette widths 0.28 and 0.19 for whole blood and leukocyte isolates respectively, P < 0.01). CONCLUSION: Our results support the use of whole blood to derive gene expression data in sepsis studies investigating novel diagnostics and subtype discovery. This strategy has a number of practical advantages, and the resulting data also have potential utility in developing molecular classifications of sepsis syndromes.
RATIONALE: Most gene expression studies of sepsis have used either whole blood or specific leukocyte fractions as source tissues for RNA. Data regarding the relative utility of these different tissue sources are lacking. OBJECTIVES: To evaluate the utility of different source tissues in studying gene expression in sepsis. METHODS: We undertook a systematic analysis of sepsis gene expression studies, including both adult and pediatric cohorts. We used clustering methods to partition samples according to gene expression levels, and compared expression cluster labels to clinical diagnoses. We also quantified the strength of cluster formation based on expression data from different tissue sources using average silhouette widths as a measure of cluster cohesiveness. RESULTS: We included 22 separate expression datasets. Whole blood was used as the source tissue in 15 studies, while leukocyte isolates were used in seven studies. Whole blood samples yielded greater specificity for the diagnosis of sepsis than data from leukocyte isolates (94% vs 78%, P = 0.03). Whole blood-derived data also yielded more cohesive clusters (median silhouette widths 0.28 and 0.19 for whole blood and leukocyte isolates respectively, P < 0.01). CONCLUSION: Our results support the use of whole blood to derive gene expression data in sepsis studies investigating novel diagnostics and subtype discovery. This strategy has a number of practical advantages, and the resulting data also have potential utility in developing molecular classifications of sepsis syndromes.