| The rapid development of computer sciences and high-throughput techniques such as Oligonucleotide and cDNA microarrays enable measuring the expression of thousands of genes simultaneously. This possibility offers an unprecedented opportunity to characterize the underlying mechanisms of a living cell. Activities of a living cell are very complex. Recently, researchers have made tremendous effort to identify coherent gene groups and to predict the gene splice sites. In this paper, we proposed some novel approaches to deal with the problems.Biological processes often contain tens of thousands of genes; the search for the gene sets involved in the same functional groups or biological pathway related to the different biological processes is still research focuses in computational biology. Nowadays, a large number of clustering methods have been used to identify the possible existence of gene groups related to different biological processes; however, majority methods, lack the attention to the overlapping relations between gene functional groups. It is the most common that the same gene involved in different biological pathway synchronously. In the paper, we use Cfinder, the overlapping community method for complex network, to identify gene groups. By analyzing the network topology of microarray data, we identify potential functional groups, and then, find gene groups involved in different biological processes. Finally, make a further in-depth analysis for the inter group overlapping genes. The results show that the gene groups we found by our method have a high functional similarity. In terms of gene enrichment analysis, each gene set has a very significant score, which indicates the genes in the same set are involved in the same biological process. In addition, the genetic overlap between the different gene groups really has extraordinary importance is the so-called focus genes. To build the more reasonable gene network, we take attention to the correlations between genes. We use Pearson coefficient and dynamic time warping (DTW) algorithm to calculate the correlations, and the results show that DTW is more efficient. In this study, we find that allowing gene groups to have overlapping is an innovative development of clustering algorithm, which is coincident with the real biological process. Based on method, we can find added genes involved in more activities, more important biological processes. Our clustering algorithm allows communities with overlap, which meet with the real biological system.Gene is a DNA sequence which carrying a multitude of genetic information, it plays a very important role in the whole life of a living cell. Therefore, identification of the location of genes in the DNA sequence is the focus of interest of many researchers. The experimental results show that there are both coding region which is called"exons"and non-coding region called"introns"in the DNA sequence. The introns were cut off in the transcription from DNA to mRNA. It is known that the junction of exon and intron is described as splice site. In this paper, the identification of the splice site is the focus of our work. We use the support vector machine to identify splice sites, and the SVM is an widespread used method for feature selection. Admittedly, applying traditional SVM on gene sequencing can provide us with new perspectives on cellular processes. However, there is a problem about the data should be highlighted: the gene sequence data we get is always unbalance. It means that the number of real splice sites is far less than the number of false. In this case, the results of traditional SVM are not precise enough. Different form the traditional method, an kind of Weighted SVM can finally solve the problem of unbalance data. The experimental results demonstrate that the WSVM achieves higher accuracy than the traditional one.Biological processes are always carried out through large numbers of genes and their products, and these activities are always complex. In this paper, we use a novel method called Cfinder to identify the gene communities. On the other hand, in order to solve the problem of unbalance data, we carry out the method of WSVM and we get a satisfactory result. |
