摘要
面对小样本海量特征的基因芯片数据,准确高效地筛选分类特征基因是数据分析的重要工作。通过制备常温状态和低温状态两类地方民猪样本的基因芯片数据,以抗寒性状相关基因筛选为目的,提出目标风险最小特征搜索算法,通过设定目标参数筛选出抗寒性状相关的基因集合。这些抗寒性状相关基因对两类样本分类效果非常好,通过与经典的分类方法相比较,多种方法都获得稳定高效的分类准确率,该特征筛选过程获得的抗寒相关基因具有较强的分类效能。通过基因功能生物学注释,得到特征基因所在的功能结点信息,这些已知基因的功能为进一步研究特征基因集合中未知基因的功能提供了试验设计依据。目标风险最小特征搜索的同时可以进行样本分类和重要分类特征的筛选,适用于基因表达谱数据的统计分析。
Facing the characteristics of small sample of gene chip data, accurate and efficient screening classification feature gene is an important work for microarray data analysis. Based on the normal tem- perature state and cold state induced pig samples, the target risk-minimized feature-searching algo- rithm （TRFA） is proposed, through the selection of target parameters, cold resistance related genes were obtained. Two samples were well classified based on these genes. Comparing with the traditional classification methods, the proposed method had good classification efficiency. Through gene function annotation, the characteristics of gene＇s function node was obtained, which provides the basis of ex- perimental design for the genes featured in the same set but their function unconfirmed. Target risk- minimized feature-searching algorithm can be used to analysis large scale gene expression profiles and finish both tasks of sample classification and main feature screening for classification.
出处
《西北农业学报》
CAS
CSCD
北大核心
2013年第10期15-21,共7页
Acta Agriculturae Boreali-occidentalis Sinica
基金
黑龙江省自然科学青年基金(QC2012C034)
黑龙江省博士后基金(LBH-Z10039)
关键词
基因芯片
特征搜索
抗寒
分类效能
目标风险最小特征算法
Microarray
Characteristics
Anti-freezing
Classification performance
Target risk-mini- mized feature-searching algorithm （TRFA）
