APOC1-Associated VLDL-Pathway Dysregulation in Papillary Thyroid Carcinoma: A Reproducible Cross-Cohort Transcriptomic Signature Associated With the Diagnostic-to-Treatment Interval.

The diagnostic-to-treatment interval (DTI) of papillary thyroid carcinoma (PTC) has long been regarded as an inactive administrative period. We expected to find that DTI is a window of high metabolic activity where apolipoprotein dysregulation may lead to harm. A four-phase integrated design was used. Phase 1 analysed transcriptomic and clinical data from TCGA-THCA (n = 209) to find DTI-associated molecular alterations, and used a maximally selected rank statistic to set the optimal time window. Phase 2 verified the lipoprotein signature in three independent GEO cohorts (GSE29265/GSE33630/GSE60542). Phase 3 used XGBoost regression in conjunction with SHAP (SHapley Additive exPlanations) to explore gene-gene interactions and predictive structures. Phase 4 used two-sample Mendelian randomization to determine whether genetically determined levels of the 233 metabolic biomarkers were associated with PTC. A 77-day threshold optimally stratified postoperative complete response rates (91.5% for DTI ≤ 77 days vs. 79.0% for DTI > 77 days; p = 0.035). Differential expression analysis found 77 genes that were significantly associated with prolonged DTI, and the enriched pathways included PPAR signalling, fat digestion and apolipoprotein particle dynamics. Linear regression showed that longer DTI was associated with higher expression of APOA1 and APOC1. XGBoost-SHAP analysis showed that APOA1 and APOC1 were the top predictive features, and their interaction explained most of the non-additive variance. Cross-cohort validation showed that APOC1 and LPL were consistently upregulated and VLDLR was downregulated in all GEO datasets. Disease-state predictive model building found that the LPL-APOC1 axis effectively distinguished tumour from normal tissue. In two-sample MR analysis of 233 metabolic traits, genetically predicted VLDL and triglyceride-rich lipoprotein traits showed a predominantly risk-increasing direction of effect, and HDL cholesterol exhibited a protective effect, which was consistent with the transcriptomic results; however, no individual exposure remained significant after multiple testing correction in either outcome dataset, so the genetic analysis provided support for and extended the hypothesis rather than confirming it. Based on the above analysis, the pre-operative period of PTC may be metabolically active rather than quiescent; however, this has not been confirmed in future studies. Dysregulation of the APOC1-associated VLDL pathway is a reproducible cross-cohort transcriptomic signature in PTC; although genetic analysis has been conducted directionally, strong causal evidence has not been obtained. Therefore, at the same time, we will shift from time-limited to metabolically targeted preoperative risk assessment.