Table Table11 provides the clinical characteristics of the derivation cohort, selleck Carfilzomib consisting of 148 patients without SSAKI and 31 patients with SSAKI. The patients with SSAKI had a higher Pediatric Risk of Mortality score and a higher mortality rate, compared with the patients without SSAKI. All other variables shown in Table Table11 were not significantly different between the two groups.Table 1Clinical characteristics of the derivation cohortIn the first derivation stage we conducted a two-step statistical test to determine which gene probes on the array (>80,000 gene probes) were differentially regulated between patients with and without SSAKI. In step one we conducted a three-group analysis of variance using normal controls (n = 53), patients without SSAKI, and patients with SSAKI as the comparison groups, and corrections for multiple comparisons (Benjamini-Hochberg false discovery rate = 5%).

This was followed by a post hoc test (Tukey) to isolate the gene probes differentially regulated between patients with and without SSAKI (100 gene probes, see Additional file 1).The 100-gene probe list presented in Additional file 1 corresponds to 61 unique and well-annotated genes. Twenty-one of the gene probes were upregulated in the patients with SSAKI, relative to the patients without SSAKI (Table (Table2).2). These 21 gene probes were subsequently used in a leave-one-out cross-validation procedure (Support Vector Machines algorithm) to predict ‘SSAKI’ and ‘no SSAKI’ classes in the derivation cohort.

The leave-one-out cross-validation procedure removes a single observation from the original sample as validation – and analyzes the remaining observations as comparators. The procedure is repeated for each observation in the sample so that each is used once as validation. Figure Figure11 provides the 2 �� 2 contingency table demonstrating the results of the leave-one-out Dacomitinib cross validation procedures, and the associated performance calculations. Figure Figure11 demonstrates that the expression patterns of these 21 upregulated gene probes can predict SSAKI with a high degree of sensitivity and modestly high specificity in the derivation cohort. In addition, the expression patterns of these 21 upregulated gene probes have a high negative predictive value for SSAKI in the derivation cohort. Accordingly these 21 gene probes represent potential candidate biomarkers for predicting SSAKI.Table 2Gene probes upregulated in patients with kidney injury that predict ‘no SSAKI’ and ‘SSAKI’ classesFigure 1Results of the leave-one-out cross-validation procedure involving 21 gene probes. The procedure was based on a Support Vector Machines algorithm and was targeted at prediction of ‘SSAKI’ and ‘no SSAKI’ classes. Performance calculations provided as the …