Processed 4D-DIA proteomic and clinicopathological data supporting a four-protein model for lymph node metastasis prediction in esophageal squamous cell carcinoma

This dataset supports a study investigating proteomic biomarkers for lymph node metastasis (LNM) in esophageal squamous cell carcinoma (ESCC). Tumor tissues from 100 patients with ESCC, including 44 patients with pathological LNM and 56 patients without LNM, were analyzed using 4D data-independent acquisition (4D-DIA) proteomics. The cohort was stratified into a development cohort of 60 patients and an independent test cohort of 40 patients. All feature selection, model development, hyperparameter optimization, calibration, and threshold determination were performed exclusively in the development cohort, whereas the test cohort was used only for one-time evaluation of the frozen model. A total of 8,614 proteins were quantified. Proteins with more than 50% missing values in the development cohort were excluded, resulting in 7,409 proteins for subsequent analysis. Candidate LNM-associated proteins were identified by covariate-adjusted differential expression analysis and ranked using a consensus feature-selection framework involving nine algorithms. Recursive feature elimination identified a four-protein biomarker panel, and Gaussian naive Bayes was selected as the final classification algorithm. The deposited materials include a processed proteomic abundance matrix, and de-identified clinicopathological data.