Detection of critical transition states in complex diseases based on distance correlation coefficient

Abstract

During the development of many complex diseases, biological systems may pass through an unstable critical transition state before disease onset or further deterioration. Timely detection of this state is important for identifying early warning signals and for supporting intervention before marked disease progression. This study proposes a model-free single-sample method based on the distance correlation coefficient, namely dCor-LNWD, for assessing disease-associated perturbations of individual diseased samples. This method uses distance correlation to evaluate both linear and nonlinear associations between gene-expression levels. This study applied dCor-LNWD to four stage-stratified cancer datasets (ESCA, KIRC, KIRP, and LUAD) from the TCGA database and the GSE13268 dataset from the GEO database, and successfully identified critical transition states in five complex disease datasets, including stage-wise critical states during cancer progression.