Confirmation of a BRAF mutation-associated gene expression signature in melanoma
Confirmation of a BRAF mutation-associated gene expression signature in melanoma
复制标题
DOI：
10.1111/j.1600-0749.2007.00375.x
复制
发表时间：
2007-06-01
期刊：
影响因子：
--
通讯作者：
Hayward, Nicholas
中科院分区：
文献类型：
--
作者：
Johansson, Peter;Pavey, Sandra;Hayward, Nicholas
BRAF癌基因的突变发生在大多数黑色素瘤中，导致促分裂原激活蛋白激酶途径的激活和下游效应子的转录。由于BRAF及其效应子可能是良好的黑色素瘤治疗靶标，因此定义了由于BRAF突变激活而受到差异调节的基因曲目是一个重要的目标。为了实现这一目标，我们和其他人试图确定是否存在与BRAF突变相关的基因表达谱。结果混合在一起，一些小组报告了BRAF签名，另一组没有。在这里，我们解决了这个问题并确认，尽管基因相关性无法揭示其表达与BRAF状态相关的特定基因，但可以通过分析全局表达模式来区分BRAF签名。具体而言，我们在这里已应用支持向量机（SVM）分析，以从63个黑色素瘤细胞系中的Affymetrix微阵列数据中进行分析。 SVM在训练样品中发现了BRAF签名，并在其余样品中以高精度（AUC = 0.840）预测BRAF突变状态。我们通过在三个已发表的微阵列数据集中重复分析来验证这是一个广义的BRAF签名，并再次发现SVM可以很好地预测BRAF突变（费城：AUC = 0.788; Zurich; Zurich：auc = 0.688; Mannheim; Mannheim：AUC：AUC = 0.686）。对我们数据训练的300个SVM的集合还预测了三个已发布的数据集中的两个（费城AUC = 0.778; Zurich AUC = 0.719; Mannheim AUC = 0.564）。综上所述，这些数据支持BRAF突变特异性表达特征的存在。
Mutations in the BRAF oncogene occur in the majority of melanomas, leading to the activation of the mitogen-activated protein kinase pathway and the transcription of downstream effectors. As BRAF and its effectors could be good melanoma therapy targets, defining the repertoire of genes that are differentially regulated because of BRAF mutational activation is an important objective. Towards this goal, we and others have attempted to determine whether a BRAF mutation-associated gene expression profile exists. Results have been mixed, with some groups reporting a BRAF-signature and another group not. Here we resolve this issue and confirm that while gene-by-gene correlations fail to reveal a specific gene(s) whose expression correlates with BRAF status, a BRAF signature can be distinguished by analysis of global expression patterns. Specifically, we have here applied support vector machine (SVM) analysis to Affymetrix microarray data from a panel of 63 melanoma cell lines. SVMs found a BRAF signature in training samples and predicted BRAF mutation status with high accuracy (AUC = 0.840) in the remaining samples. We verified this is a generalized BRAF signature by repeating the analysis in three published microarray datasets, and again found that SVMs predicted BRAF mutation well (Philadelphia: AUC = 0.788; Zurich: AUC = 0.688; Mannheim: AUC = 0.686). An ensemble of 300 SVMs trained on our data also predicted BRAF mutation status in two of the three published datasets (Philadelphia AUC = 0.778; Zurich AUC = 0.719; Mannheim AUC = 0.564). Taken together, these data support the existence of a BRAF mutation-specific expression signature.