Background: Tumor heterogeneity imposes great challenges on cancer treatment. Cancer stem cells (CSCs) are a leading factor contributing to tumor occurrence. However, the mechanisms underlying the growth of thyroid cancer (TCHA) are still unclear.

Methods: Key genes regulating the characteristics of THCA, such as stemness were identified by combining gene expressions of samples downloaded from the Cancer Genome Atlas (TCGA) and were used to establish an mRNA expression stemness index (mRNAsi) through machine learningbased methods. The relationships of mRNAsi, THCA clinical features and molecular subtypes were analyzed. Weighted Gene Co-Expression Network Analysis (WGCNA) was performed to obtain mRNAsi-related gene modules and determine mRNAsi-related differentially co-expressed genes. Key genes related to mRNAsi were screened by protein interaction network. Functional analysis was conducted and expressions of key genes were verified in multiple external data sets.

Results: The mRNAsi score, which was found to be lower in the TCHA tissues than that in normal tissues (p<0.05), was positively correlated with a slow progression of tumor prognosis (p=0.0085). We screened a total of 83 differentially co-expressed genes related to mRNAsi and multiple tumor pathways such as apoptosis, tight junction, cytokine-cytokine receptor interaction, and cAMP signaling pathway (p<0.05). Finally, 28 protein interaction networks incorporating 32 genes were established, and 3 key genes were identified through network mining. 3 core genes were finally determined, as their low expressions were strongly correlated with the progression of THCA.

Conclusion: The study found that NGF, FOS, and GRIA1 are closely related to the characteristics of THCA stem cells. These genes, especially FOS, are highly indicative of the prognosis of THCA patients. Thus, screening therapy could be used to inhibit the stemness of TCHA.

背景：肿瘤异质性给癌症治疗带来了巨大挑战。癌症干细胞 (CSC) 是导致肿瘤发生的主要因素。然而，甲状腺癌（TCHA）生长的潜在机制仍不清楚。

方法：通过结合从癌症基因组图谱（TCGA）下载的样本的基因表达，确定调控 THCA 特征的关键基因，如干性，并通过基于机器学习的方法建立 mRNA 表达干性指数（mRNAsi）。分析mRNAsi、THCA临床特征和分子亚型的关系。进行加权基因共表达网络分析（WGCNA）以获得与mRNAsi相关的基因模块并确定与mRNAsi相关的差异共表达基因。通过蛋白质相互作用网络筛选与mRNAsi相关的关键基因。进行了功能分析，并在多个外部数据集中验证了关键基因的表达。

结果：发现TCHA组织中mRNAsi评分低于正常组织(p<0.05)，与肿瘤预后的缓慢进展呈正相关(p=0.0085)。我们共筛选了 83 个与 mRNAsi 和多种肿瘤通路（如凋亡、紧密连接、细胞因子-细胞因子受体相互作用和 cAMP 信号通路）相关的差异共表达基因（p<0.05）。最终建立了包含32个基因的28个蛋白质相互作用网络，并通过网络挖掘确定了3个关键基因。最终确定了 3 个核心基因，因为它们的低表达与 THCA 的进展密切相关。

结论：研究发现NGF、FOS、GRIA1与THCA干细胞的特性密切相关。这些基因，尤其是 FOS，高度指示 THCA 患者的预后。因此，筛选疗法可用于抑制TCHA的干性。