Association between AUP1 expression and lipid-rich phenotypes in clear cell renal cell carcinoma: a CT-based radiogenomic study

Abstract Objectives Clear cell renal cell carcinoma (ccRCC) is characterized by abnormal lipid accumulation, which may influence tumor behavior and imaging appearance. However, the relationship between lipid metabolism–related genes, such as ancient ubiquitous protein 1 ( AUP1 ), and computed tomography (CT) imaging features remains unclear. This study aims to investigate the association between AUP1 expression and CT features in ccRCC patients, using artificial intelligence (AI)-based radiogenomic analysis. Methods Genomic and imaging data from 208 patients were analyzed. Regions of interest (ROIs) were manually placed on solid tumor tissue, excluding necrotic areas. An AI-based workflow, implemented in Python with the scikit-learn library, extracted first-order radiogenomic features, median, minimum, and maximum Hounsfield unit (HU) values, as indirect surrogates of intracellular lipid content, with lower HU values indicating higher lipid accumulation. Distributional descriptors were also computed to capture intra-tumoral heterogeneity. Associations between imaging features, AUP1 expression, and histological grade were tested using non-parametric statistics (Wilcoxon, Kruskal-Wallis tests, Pearson’s Chi-square test, and Fisher’s exact test, p