Co-occurrence and permutation-based analysis pipeline for syndrome-element network transition across gastric precancerous lesion stages

The pipeline separates prevalence drift from relation reorganization using a marginal-frequency-corrected lift measure, patient-level stage-label permutation tests with Benjamini-Hochberg control, weighted network topology, a directional drift statistic D that operationalizes the classical "qi disorder progressing into blood" transmission, counterfactual cohorts that vary one mechanism at a time, an error-propagation stress test, and a TreeSHAP interaction module whose decision rule is calibrated against 1,000 label permutations. This study uses simulated data only; no human participants and no personal information are involved.

Every number in the associated manuscript regenerates from the global random seed 20260913. The package includes the analysis code, the simulated cohort, the data-generating-process parameters with seeds, the syndrome-element mapping table, the keyword rule table, and all edge-level results.