The application of artificial intelligence methods to gene expression data for differentiation of uncomplicated and complicated appendicitis in children and adolescents - a proof of concept study –

Abstract

Background: Genome wide gene expression analysis has revealed hints for independent immunological pathways underlying the pathophysiologies of phlegmonous (PA) and gangrenous appendicitis (GA). Methods of artificial intelligence (AI) have successfully been applied to routine laboratory and sonographic parameters for differentiation of the inflammatory manifestations. In this study we aimed to apply AI methods to gene expression data to provide evidence for feasibility. Methods: Modern algorithms from AI were applied to 56.666 gene expression data sets from 13 patients with PA and 16 with GA aged 7–17 years by using resampling methods (bootstrap). Performance with respect to sensitivities and specificities where investigated with receiver operating characteristic (ROC) analysis. Results: Within the experimental setting a best performing discriminatory biomarker signature consisting of a set of 4 genes could be defined: ERGIC and golgi 3, regulator of G-protein signaling 2, Rho GTPase activating protein 33, and Golgi Reassembly Stacking Protein 2. ROC analysis showed a mean area under the curve of 84%. Conclusions: Gene expression based application of AI methods is feasible and represents a promising approach for future discriminatory diagnostics in children with acute appendicitis.