Abstract:
Machine learning (ML) algorithms were used to identify a novel biological target for breast cancer and explored its relationship with the tumor microenvironment (TME) and patient prognosis. The edgR package identified hub genes associated with overall survival (OS) and prognosis, which were validated using public datasets. Of 149 up-regulated genes identified in tumor tissues, three ML algorithms identified COL11A1 as a hub gene. COL11A1was highly expressed in breast cancer samples and associated with a poor prognosis, and positively correlated with a stromal score (r=0.49, p<0.001) and the ESTIMATE score (r=0.29, p<0.001) in the TME. Furthermore, COL11A1 negatively correlated with B cells, CD4 and CD8 cells, but positively associated with cancer-associated fibroblasts. Forty-three related immune-regulation genes associated with COL11A1 were identified, and a five-gene immune regulation signature was built. Compared with clinical factors, this gene signature was an independent risk factor for prognosis (HR=2.591, 95%CI 1.831-3.668, p=7.7e-08). A nomogram combining the gene signature with clinical variables, showed better predictive performance (C-index=0.776). The model correction prediction curve showed little bias from the ideal curve. COL11A1 is a potential therapeutic target in breast cancer and may be involved in the tumor immune infiltration; its high expression is strongly associated with poor prognosis.
摘要:
机器学习(ML)算法用于识别乳腺癌的新生物靶标,并探索其与肿瘤微环境(TME)和患者预后的关系。edgR软件包确定了与总生存期(OS)和预后相关的hub基因,使用公共数据集进行了验证。在肿瘤组织中鉴定出的149个上调基因中,三种ML算法将COL11A1识别为hub基因。COL11A1在乳腺癌样本中高表达,并与不良预后相关。并与TME中的基质评分(r=0.49,p&lt;0.001)和ESTIMATE评分(r=0.29,p&lt;0.001)呈正相关。此外,COL11A1与B细胞呈负相关,CD4和CD8细胞,但与癌症相关的成纤维细胞呈正相关。鉴定了43个与COL11A1相关的免疫调节基因,建立了5个基因的免疫调节标志。与临床因素相比,该基因标记是影响预后的独立危险因素(HR=2.591,95CI1.831-3.668,p=7.7e-08).将基因签名与临床变量相结合的列线图,表现出更好的预测性能(C指数=0.776)。模型校正预测曲线显示与理想曲线的偏差很小。COL11A1是乳腺癌潜在的治疗靶点,可能参与肿瘤的免疫浸润,其高表达与预后不良密切相关。