4.2. Data Generation

The transcriptomic discovery dataset was generated from whole-blood samples obtained from KD patients, healthy controls, and patients with bacterial and viral infections using the HumanHT-12 version 4.0 (Illumina, San Diego, CA, USA) microarray [18]. In order to obtain a transcriptomic validation dataset containing the same disease groups as the transcriptomic discovery dataset, two datasets were merged. One dataset contained gene expression values (HumanHT-12 version 4.0 [Illumina, San Diego, CA, USA] microarray) from whole-blood samples obtained from acute and convalescent KD samples [19]. The other dataset consisted of gene expression values (HumanHT-12 version 3.0 [Illumina, San Diego, CA, USA] microarray) from whole-blood samples obtained from patients with bacterial and viral infections [22]. For all three independent microarray experiments, one batch of samples was processed, and samples were randomly positioned across the arrays.

The proteomic discovery dataset was generated from plasma samples using LC–MS/MS. Full details of the experimental protocol are in the Supplementary Text. The proteomic validation dataset was generated from serum samples using the SomaScan (SomoLogic, Boulder, CO, USA) aptamer-based platform [20]. Prior to pre-processing, 867 proteins were measured in the discovery dataset (LC–MS/MS) and 1300 in the validation dataset (SomaScan). Samples in the proteomic validation dataset were split across three plates with KD, DB, DV and HC samples present on each plate in relative proportions.