Plasma autoantibody profiling via PhIP-seq and machine learning for diagnosis of esophageal squamous cell carcinoma

Early diagnosis of esophageal squamous cell carcinoma (ESCC) is crucial for improving patient survival. This study employed phage immunoprecipitation sequencing (PhIP-Seq) to comprehensively profile circulating autoantibodies using 1039 plasma samples from a multicenter cohort. Through integrated machine learning approaches, we identified a panel of core antigenic peptides that effectively distinguish ESCC patients from normal controls. The established diagnostic model, based on XGBoost and featuring autoantibodies against antigens such as KMT2C and INPP4B, demonstrated robust performance across multiple validation sets. It achieved area under the curve (AUC) values of 0.912 (95% CI: 0.878-0.946) in the training set, 0.782 (95% CI: 0.696-0.868) in the internal validation set, 0.809 (95% CI: 0.762-0.856) and 0.729 (95% CI: 0.570-0.889) in two independent external cohorts, respectively. Furthermore, the model showed preliminary potential in assessing treatment response, suggesting its possible utility in disease monitoring, although this finding requires independent validation in larger prospective cohorts. This work provides a framework for non-invasive ESCC detection and precision management. Oesophageal squamous cell carcinoma is often diagnosed in later stages, which limits potential treatment options. Here, the authors develop a PhIP-seq model to diagnose oesophageal squamous cell carcinoma across multiple cohorts.