Merging Temporal Changes in Clinical Informatics, Transcriptomics, and Cytokine Profiles to Understand the Host Response to Bacteremia

点击翻译按钮获取中文摘要

As an ICU physician and an immunologist, I have devoted my research career to understanding sepsis, a disease that affects nearly 2 million people annually in the USA. Sepsis is a life-threatening condition that arises when the body's response to infection injures its own tissues. Great strides have been made towards improving the clinical care of the septic patient, but the mortality rate remains >20% for several reasons: First, each sepsis-causing pathogen, be it bacterial, viral, or fungal, carries its own virulence factors that affect the host response, but unfortunately, much research has focused on studying septic patients as a group, rather than distinguishing patients based on microbiologic cause. Second, researchers often study patients once they manifest sepsis-induced organ failure yet many people sustain infections due to common sepsis pathogens, like Staphylococcus aureus, but never develop sepsis; understanding the "appropriate" response to a pathogen is critical to understanding the "inappropriate" response of sepsis. Finally, much sepsis research occurs in silos; clinician researchers focus on the electronic medical record, while basic scientists analyze biologic data. Too often, these groups do not collaborate to share information, even though understanding the biologic significance of clinical data may be of great value. To address these issues, I propose a unique approach to understand the host response to infection that incorporates both biologic data and clinical electronic medical record (EMR) data from patients with S. aureus bacteremia. By limiting analysis of the host response to infections caused by a single pathogen at a single site, we can control for the variability induced by pathogen-specific factors. In addition, by studying all patients with S. aureus bacteremia, and not simply those patients with sepsis, we can understand both the appropriate host response as well as the inappropriate host response that characterizes the development of sepsis. With our bank of samples collected from S. aureus bacteremia patients we will analyze both cellular mRNA transcripts and plasma protein/cytokine levels, collected at different time points from each patient. We will combine the results of these analyses with clinical data found in the EMR to provide a correlation between the biology of the host response and its clinical manifestations. Our per-patient data, then, will have unprecedented granularity which we can then use to apply machine learning techniques to identify multi-faceted endotypes that predict outcomes (such as mortality). Once we have built these endotype models, we will validate them using pilot data collected from newly enrolled patients with either S. aureus or E. coli bacteremia. This approach will allow us to identify factors common to the dysregulated host response across all infections, as well as those that may be specific to the type of infection. Understanding both the appropriate and the inappropriate host response to infection, and understanding which aspects of the host response are pathogen-specific (and which are not) will allow development of novel therapies for this devastating disease.