Machine learning‐driven diagnostic signature provides new insights in clinical management of hypertrophic cardiomyopathy

Abstract

AbstractAimsIn an era of evolving diagnostic possibilities, existing diagnostic systems are not fully sufficient to promptly recognize patients with early‐stage hypertrophic cardiomyopathy (HCM) without symptomatic and instrumental features. Considering the sudden death of HCM, developing a novel diagnostic model to clarify the patients with early‐stage HCM and the immunological characteristics can avoid misdiagnosis and attenuate disease progression.Methods and resultsThree hundred eighty‐five samples from four independent cohorts were systematically retrieved. The weighted gene co‐expression network analysis, differential expression analysis (|log2(foldchange)| > 0.5 and adjusted P < 0.05), and protein–protein interaction network were sequentially performed to identify HCM‐related hub genes. With a machine learning algorithm, the least absolute shrinkage and selection operator regression algorithm, a stable diagnostic model was developed. The immune‐cell infiltration and biological functions of HCM were also explored to characterize its underlying pathogenic mechanisms and the immune signature. Two key modules were screened based on weighted gene co‐expression network analysis. Pathogenic mechanisms relevant to extracellular matrix and immune pathways have been discovered. Twenty‐seven co‐regulated genes were recognized as HCM‐related hub genes. Based on the least absolute shrinkage and selection operator algorithm, a stable HCM diagnostic model was constructed, which was further validated in the remaining three cohorts (n = 385). Considering the tight association between HCM and immune‐related functions, we assessed the infiltrating abundance of various immune cells and stromal cells based on the xCell algorithm, and certain immune cells were significantly different between high‐risk and low‐risk groups.ConclusionsOur study revealed a number of hub genes and novel pathways to provide potential targets for the treatment of HCM. A stable model was developed, providing an efficient tool for the diagnosis of HCM.