A novel combinatorial score for feature selection with P-Tree in DNA microarray data analysis

DNA microarray experiments are being used to gather information from tissue and cell samples by generating thousands of gene expression measurements. Many researchers are conducting researches regarding gene expression differences, which is useful in disease diagnose, outcome prediction, cancer type classification and etc. In mining high-dimensional microarray data, feature selection is an important pre-processing stage. In the literature nearly all existing supervised feature selection methods use class labels as supervision information. In this paper, we propose a novel score using the label correlation in combination with the correlation between features. We design a Combinatorial Score feature selection algorithm in P-Tree structure and combine it with K-Nearest-Neighbor algorithm for breast cancer clinic metastasis time prediction. Our experiments suggest that our Combinatorial Score feature selection algorithm can find a subset of genes with high computation efficiency and significant performance for breast cancer clinical metastasis prediction.