A computational workflow for the detection of candidate diagnostic biomarkers of Kawasaki disease using time-series gene expression data

摘要: Abstract Unlike autoimmune diseases, there is no known constitutive and disease-defining biomarker for systemic autoinflammatory diseases (SAIDs). Kawasaki disease (KD) one of the “undiagnosed” types SAIDs whose pathogenic mechanism gene mutation still remain unknown. To address this issue, we have developed a sequential computational workflow which clusters KD patients with similar expression profiles across three different phases (Acute, Subacute Convalescent) utilizes resulting clustermap to detect prominent genes that can be used as diagnostic biomarkers KD. Self-Organizing Maps (SOMs) were employed cluster expressions through inter-phase intra-phase clustering. Then, false discovery rate (FDR)-based feature selection was applied significantly deviate per-phase clusters. Our results revealed five candidate diagnosis, namely, HLA-DQB1, HLA-DRA, ZBTB48, TNFRSF13C, CASD1. our knowledge, these are reported first time in literature. The impact discovered diagnosis against ones demonstrated by training boosting ensembles (AdaBoost XGBoost) classification on common platform cross-platform datasets. classifiers trained proposed from data yielded an average increase 4.40% accuracy, 5.52% sensitivity, 3.57% specificity than Acute phases, followed notable 2.30% 2.20% 4.70% analysis.