The molecular sepsis signature*
In general, reliability and reproducibility of transcriptomics data are encouraging to date (2). Contrary to a rather disappointingly slow progress of diagnostic (plasma) proteomics, the recent past has seen intriguing developments in transcriptomics, in particular, toward cancer diagnostics.
However, although several commercial products like MammaPrint (Agendia, Huntington Beach, CA) or OncoType DX (Genomic Health, Redwood City, CA) have been launched lately for applications in cancer diagnostics, sepsis remains a special challenge for several obvious reasons. Sepsis, by definition, includes an infection for which clear evidence is still a problem, leave alone the reliable identification of the pathogen (3, 4). Conventional blood culture assays are notoriously unreliable and biased by antibiotics treatment; recent polymerase chain reaction–based assays are not yet established and validated for clinical routine application. Furthermore, polymerase chain reaction assays are prone to contaminations.
Hence, the special problem here is that it is hard to establish a reliable set of clinical reference samples with safe diagnosis and unlike other diseases, there is no straightforward independent assay such as immunohistologic confirmation after surgery.
In light of all these difficulties, the attempt to more precisely characterize sepsis on the molecular level is still a brave endeavor. Being aware of these principal problems, the authors of the article tried to provide evidence for infection in the samples subjected to microarray analysis (1). Unfortunately, the cohort was heterogeneous and the evidence for infection was partially insufficient. These clearly are weak points of the study as well as the total size of the cohort. Given the problems mentioned above, it is not realistic to expect valid molecular signatures at numbers below several hundreds unless the patients are homogeneous and carefully preselected. However, although the authors did not provide means for improved diagnosis at the moment, it has to be acknowledged that they were among the first to publish data demonstrating the usefulness of gene expression signatures to the characterization of such a complex disease as sepsis.