Abstract O.29: Transcriptional Profiling Discriminates Complete and Incomplete KD from Human Adenovirus

Abstract

Background: The diagnosis of Kawasaki disease (KD) is often difficult to distinguish from HAdV. Objective: 1) To characterize the specific transcriptional profiles of KD patients versus acute HAdV infection 2) To determine whether the molecular distance to health (MDTH) score (a molecular score that reflects the perturbation derived from whole genome transcriptional analysis) correlates with response to therapy. Methods: Whole blood RNA samples collected in Tempus tubes were analyzed using Illumina chips and GeneSpring software 7.4 from 76 pediatric patients with complete KD, 13 with incomplete KD, and 19 patients with HadV, and 20 age- and sex-matched healthy controls (HC). We used class comparison algorithms (Mann-Whitney p< 0.01, Benjamini-Hochberg, and 1.25- fold change filter) and modular analysis to define the KD profiles; class prediction algorithm was used to identify genes that best differentiate KD and HAdV. Results: Statistical group comparisons identified 7,899 genes differentially expressed in 39 complete KD patients versus HC (KD biosignature). This signature was validated in another 37 patients with complete KD and in 13 patients with incomplete KD. Modular analysis in children with complete KD demonstrated overexpression of inflammation, neutrophils, myeloid cell, coagulation cascade, and cell cycle genes. The class prediction algorithm identified 25-classifier genes that differentiated children with KD vs HAdV infection in two independent cohorts of patients with 92% (95% CI [73%-99%]) sensitivity and 90% [67%-98%] specificity. MDTH scores in KD patients significantly correlated with the baseline c-reactive protein (R=0.29, p=0.008) and was four fold higher than in children with HAdV (p<0.01). In addition, KD patients that remained febrile 36 hours after treatment with IVIG (non-responders) demonstrated higher baseline, pre-treatment MDTH values compared with responders [12,290 vs. 5572 respectively; p=0.009]. Conclusion: Transcriptional signatures can be used as a tool to discriminate between KD and HAdV infection, and may also provide prognostic information.