response theory (1RT), human-computer interaction (HC1). A Hybrid Filter and Wrapper Algorithms for Genetic Pathway Analysis of Ovarian Cancer Microarray Dataset

摘要

Ovarian Cancer does not have a high rate of infection among all gynecological cancers; however, it has the highest mortality. A tumor marker CA-125 is used as the tool of screening presently, yet, it lacks accuracy. Finding better screening tools for ovarian cancer is thus a crucial issue. This study aims to create a model capable of identifying the stages of ovarian cancer with Microarray Dataset for its patients and by the Data Mining method, also to identify the Target Genes. The study began by choosing the key genes from the Microarray Dataset for ovarian cancer using a Filter technique which employs ANOVA and Information Gain. The next technique used was a Wrapper, which selects Target Genes from among Important Genes; the Wrapper is mainly used Bio-inspired Algorithms, including BPNN and GA. Our findings suggested that using the optimal identifier of ovarian cancer stages to test on only seven target genes can identify any of the four stages of ovarian cancer with accuracy rate at 94.63%, where the target genes are 1TGB2, PLEC, POSTN, C1S, FN 1, CDK5 and PAPPA. Such findings can serve as important biological and medical references, which, when testified by biological and medical experiments, are expected to be used as a tool of screening ovarian cancer enabling doctors to diagnose and treat it as early as possible, increasing the survival rate in ovarian cancer patients.