2885. A Host Transcriptional Signature for Accurate Diagnosis of Candidemia in the Hospital Setting

Abstract

BackgroundCandidemia is one of the most common nosocomial bloodstream infections in the United States and causes significant morbidity and mortality in hospitalized patients. Improved rapid diagnostics capable of differentiating candidemia from other causes of febrile illness in the hospitalized patient are of paramount importance. Pathogen class-specific biomarker-based diagnostics such as those focusing on host gene expression patterns in circulating leukocytes may offer a promising alternative.MethodsRNA sequencing was performed on peripheral blood samples from 27 hospitalized patients with blood culture positive invasive candidiasis. Samples from healthy controls as well as at-risk subjects with acute febrile illness and similar clinical backgrounds but other infectious or noninfectious etiologies were used as comparator phenotypes (35 subjects with culture-proven bacterial infection, 49 with confirmed viral infection, and 17 with acute noninfectious illness). Bayesian techniques were utilized to develop infection-specific classifiers and leave one out cross-validation was used to estimate the predictive probability of each pathogen class.ResultsCandidemia triggers a unique, robust and conserved transcriptomic response in human hosts with 1,170 genes differentially upregulated compared with healthy controls. Based on this strength of signal, we developed a transcriptomic classifier that was capable of identifying candidemia, viral, or bacterial infection with a high degree of accuracy (auROC for Candida 0.93, Bacterial 0.98, Viral 0.99). The Candida component of this classifier (29-genes) was able to diagnose candidemia with a sensitivity of 88% and specificity of 100%.ConclusionThe host transcriptomic response during candidemia in hospitalized adults is highly conserved and unique from the genomic responses to acute viral and bacterial infection. This approach shows promise for the development of host response-based classifiers capable of differentiating multiple types of clinical illnesses at once in at-risk febrile patients.DisclosuresEphraim L. Tsalik, MD MHS PhD, Immunexpress: Consultant; Predigen, Inc.: Officer or Board Member, Research Grant.