Abstract
Current screening methods for esophageal squamous cell carcinoma (ESCC) face challenges such as low patient compliance and high costs. This study aimed to develop a model based on oral microbiome data for identifying ESCC. By analyzing 249 oral flora samples, we identified microbial markers associated with ESCC and constructed random forest classifiers that distinguished patients with ESCC from controls, achieving an area under the ROC curve (AUC) of 0.87. Key ESCC-associated microbial markers included Neisseria perflava and Haemophilus parainfluenzae. The classifier was validated within the cohort, attaining an AUC of 0.93. For comparison, traditional tumor markers carcinoembryonic antigen (CEA) and squamous cell carcinoma antigen (SCC-Ag) yielded AUCs of 0.84. Functional analysis identified pathways linked to ESCC, such as glycerol degradation and nitrate reduction. This study suggests a potential noninvasive method for detecting ESCC, offering a more accessible and accurate alternative to current screening methods.