Exploratory Machine Learning and Omics Integration in the Search for Biomarkers of Papillary Thyroid Cancer

Papillary thyroid carcinoma (PTC) is among the most common endocrine malignancies worldwide, and although generally associated with a favorable prognosis, a subset of patients develops aggressive disease with higher recurrence risk. This highlights the need for improved molecular characterization. Data integration approaches combined with computational methods offer new opportunities to refine diagnosis and uncover disease mechanisms. This study aims to integrate omics data and apply machine learning (ML) to identify clinically relevant biomarkers in papillary thyroid carcinoma. We selected 11 genes from the differentially expressed genes (DEGs)–LASSO intersection approach. Genes were validated using an independent external dataset (AUC = 91%, Sens. = 92%, Spec. = 97%, and Acc. = 95%). DEGs were integrated with metabolomics data from the literature, enabling the construction of a metabolite–gene interaction network, highlighting norepinephrine, arachidonic acid, and glutamic acid as representative metabolites, while the main genes were SLC6A14, ADK, ATIC, NT5E, and AR. We identified potential drug–gene interactions and performed survival analysis to assess the relevance of the possible biomarkers. This novel pipeline combining integration and machine learning provides new insights into thyroid cancer biology and identifies promising diagnostic markers, supporting advances in precision medicine.