Multi-omics clustering analysis carries out the molecular-specific subtypes of thyroid carcinoma: implicating for the precise treatment strategies.

Abstract

Thyroid cancer (TC) is the most prevalent endocrine malignancy worldwide. This study aimed to explore the molecular subtypes and improve the selection of targeted therapies. We used multi-omics data from 539 patients with DNA methylation, gene mutations, mRNA, lncRNA, and miRNA expressions. This study employed consensus clustering algorithms to identify molecular subtypes and used various bioinformatics tools to analyze genetic alterations, signaling pathways, immune infiltration, and responses to chemotherapy and immunotherapy. Two prognostically relevant TC subtypes, CS1 and CS2, were identified. CS2 was associated with a poorer prognosis of shorter progression-free survival times (P < 0.001). CS1 exhibited higher copy number alterations but a lower tumor mutation burden than CS2. CS2 exhibited activation in cell proliferation and immune-related pathways. Drug sensitivity analysis indicated CS2's higher sensitivity to cisplatin, doxorubicin, paclitaxel, and sunitinib, whereas CS1 was more sensitive to bicalutamide and FH535. The different activated pathways and sensitivity to drugs for the subtypes were further validated in an external cohort. Twenty-four paired tumors and adjacent normal tissues by immunohistochemical staining further demonstrated the prognostic value of CXCL17. In conclusion, we identified two distinct molecular subtypes of TC with significant implications for prognosis, genetic alterations, pathway activation, and treatment response.