| Date |
March 9, 2010 |
| Speaker |
Dr. Narayan Behera, Centre for Ecological Research, Kyoto University |
| Title |
An evolutionary data mining algorithm to find the candidate genes from microarray gene expression data
|
| Abstract |
Clustering and gene selection from microarray gene expression data have gained tremendous importance in recent times as they help in gaining insight into the dynamics of biological systems. Gene clustering enables identification of co-expressed and co-regulated genes. Gene selection is an important aspect of data mining and analysis. It helps in identifying the genes that play pivotal roles in specified biological conditions that lead to diseased states. Today many algorithms exist for extracting these information but all have inherent limitations. In this paper a novel algorithm has been proposed for gene clustering, feature selection and classification of test samples with a higher accuracy. This algorithm is a hybrid of clustering algortihm and evolutionary computation. The evolutionary computation uses a genetic algorithm that utilizes the biological principles of evolution to solve an optimization problem in the gene clustering space. The genetic distance measure employed here is based on mutual information which takes into account similarity of the gene expression levels as well as positive and negative correlations between the genes while clustering them. A study on the gastric cancer, the colon cancer and the brain cancer microarray gene expression datasets and comparison with some existing algorithms show the present algorithm to be superior than most other conventional algorithms. This algorithm is used to find the top ranking genes that qualify to contain the most diagnostic information for the gastric cancer.
|
|
アブストラクト