癌症基因表达谱挖掘中的特征基因选择算法GA/WV

GA/WV application in tumor gene expression profiling mining as a feature selection algorithm

鉴定癌症表达谱的特征基因集合可以促进癌症类型分类的研究,这也可能使病人获得更好的临床诊断?虽然一些方法在基因表达谱分析上取得了成功,但是用基因表达谱数据进行癌症分类研究依然是一个巨大的挑战,其主要原因在于缺少通用而可靠的基因重要性评估方法。GA/WV是一种新的用复杂的生物表达数据评估基因分类重要性的方法,通过联合遗传算法(GA)和加权投票分类算法(WV)得到的特征基因集合不但适用于WV分类器,也适用于其它分类器?将GA/WV方法用癌症基因表达谱数据集的验证,结果表明本方法是一种成功可靠的特征基因选择方法。

Identification of gene subsets from gene expression analysis is useful in tumor types classifying , and it would also helps pa- tients to accept better clinic diagnosing. Though some methods for analyzing microarray expression data have shown advantages, it is still a big challenge to connect cancer classification to gene expression profiling data. In this study, we described a novel algorithm for assessing the importance of genes for sample classification based on complex expression data ： GA/WV. The gene sets we get by combi- ning genetic algorithm （GA） and weight voting （WV） methods are not only suited for WV classification but other classifictaion. We applied this GA/WV analysis to a set of gene expression data from different tumor tissues, the results of clustering and testing demon- strated that this novel algorithm is an advanced feature gene selection algorithm in gene expression data mining.