Identification of novel characteristic biomarkers and immune infiltration profile for the anaplastic thyroid cancer via machine learning algorithms.

PURPOSE

Anaplastic thyroid cancer (ATC) is a rare and lethal malignant cancer. In recent years, the application of molecular-driven targeted therapy and immunotherapy has markedly improved the prognosis of ATC. This study aimed to identify characteristic genes for ATC diagnosis and revealed the role of ATC characteristic genes in drug sensitivity and immune cell infiltration.

METHODS

We downloaded ATC RNA-sequencing data from the GEO database. Following the combination and normalization of the dataset, we first divided the combined datasets into the training cohort and the validation cohort. We identified differentially expressed genes (DEGs) in ATC by differential expression analysis in the training cohort. We used two machine learning algorithms, least absolute shrinkage and selection operator (LASSO) and support vector machine-recursive feature elimination (SVM-RFE) to identify ATC characteristic genes. The CIBERSORT algorithm was performed to calculate the abundance of various immune cells in ATC. Finally, we validated the expression of ATC characteristic genes by quantitative RT-PCR (RT-qPCR) in ATC cell lines and immunohistochemistry (IHC).

RESULTS

A total of 425 DEGs were identified in the training cohort, including 240 upregulated genes and 185 downregulated genes. Four ATC characteristic genes (ADM, PXDN, MMP1, and TFF3) were identified, and their diagnostic value was validated in the validation cohort (AUC in ROC analysis > 0.75). We established a practical gene expression-based nomogram to accurately predict the probability of ATC. We also found that ATC characteristic biomarkers are associated with the tumor immune microenvironment and drug sensitivity.

CONCLUSION

ADM, PXDN, MMP1, and TFF3 might serve as potential ATC diagnostic biomarkers and may be helpful for ATC molecular targeted therapy and immunotherapy.