结肠癌患者诊断的基因标志物识别算法*

为了得到具有强分类信息的极少结肠癌特征基因,实现对结肠癌患者的准确识别,文中提出结肠癌患者诊断的基因标志物识别算法.首先提出基因密度和基因距离的概念,构造以基因密度和基因距离分别为横纵坐标的基因2D空间散列图,选择处于密度峰值点的基因构成优选基因子集,然后采用密度峰值K中心点(DP_K-medoids)算法对降维后的结肠数据集样本进行聚类分析.基因距离和样本距离分别采用欧氏距离、曼哈顿距离、切比雪夫距离和夹角余弦距离度量.实验表明,在夹角余弦距离下,文中算法可以选择到具有高准确率、高灵敏度、高特异度和高马修斯相关系数的规模较小的结肠癌基因子集.

Gene Markers Identification Algorithm for Detecting Colon Cancer Patients

To detect those few informative genes with strong classification information and identify colon cancer patients as correctly as possible, an algorithm is proposed in this paper to identify the gene markers for detecting colon cancer patients. The densities and distances are defined for genes firstly. All genes are scattered in a 2D space with gene density and distance as X-axis and Y-axis, respectively. Those genes at high density peaks are selected to construct the optimal gene subset. Then, those samples only with genes in the optimal gene subset of colon dataset are clustered by DP_K-medoids clustering algorithm. The distances between genes or samples are calculated via Euclidean distance, Manhattan distance, Chebyshev distance and the cosine distance, respectively. The experimental results demonstrate that the proposed algorithm can find the optimal gene subset of colon cancer with high accuracy, sensitivity, specificity and MCC, and with a very few number of genes as well.