The Analysis of Microarray Datasets Using a Genetic Programming
The Analysis of Microarray Datasets Using a Genetic Programming
Chun-Gui Xu, Kun-Hong Liu, and De-Shuang Huang, Senior Member, IEEE
Abstract—Microarray technology has been widely applied to search for biomarkers of diseases, diagnose diseases and analyze gene regulatory network. Abundance of expression data from microarray experiments are processed by informatics tools, such as Supporting Vector Machines (SVM), Artificial Neural Network (ANN), and so on. These methods achieve good results in single dataset. Nevertheless, most analyses of microarray data are only focused on a series of data obtained from the same lab or gene chip. Then the discoveries may only be suitable for data they experimented on but lack of general sense. In this paper, we propose a genetic programming (GP) based approach to analyze microarray datasets. The GP implements classification and feature selection at the same time. To validate the significance of the selected genes and generated classification rules, the results are tested on different datasets obtained from different experimental conditions. The results confirm the efficiency of GP in the classification of different samples.
icroarray technology has been proposed and widely used for many years. Through simultaneous assessments of gene expressions in specific kinds of cells, the immediate temporal and spacial properties of them can be recorded [1]. Different cells have their distinct expression values of genes and it’s supposed that cells from the same tissue and in the same condition have a common expression pattern. So it’s meaningful to apply microarray to diagnose diseases, even before accumulation of abnormal gene expressions result in detectable changes in human body. Due to the high expenditure of microarray experiment and the fact that only a small portion of genes on microarray chips contribute to the disease, it’s better to sift out meaningful genes for diagnosis.
Recent years, a number of pattern recognition approaches have been proposed for selection of meaningful genes, such as Supporting Vector Machines (SVM) [2], Artificial Neural Network (ANN) [3], Bayesian Network (BN) [4], and so on. Most researchers carried out their experiments only on unique dataset obtained in the same lab or gene chip. However, in
This work was supported by the grants of the National Science Foundation of China, Nos. 30570368 & 30700161, the grants from the National High Technology Research and Development Program of China (863 Program), Nos. 2007AA01Z167 & 2006AA02Z309.
Chun-Gui Xu is with Intelligent Computing Lab, Hefei Institute of Intelligent Machines, Chinese Academy of Sciences, P.O. Box 1130, Hefei, Anhui, 230031, China (phone: 086-0551-5592751; fax: 086-0551-5592751; e-mail: xcg1984@mail.ustc.edu.cn).
Kun-Hong Liu was with Intelligent Computing Lab, Hefei Institute of Intelligent Machines, Chinese Academy of Sciences..
De-Shuang Huang is with Intelligent Computing Lab, Hefei Institute of Intelligent Machines, Chinese Academy of Sciences, P.O. Box 1130, Hefei, Anhui, 230031, China (e-mail: dshuang@iim.ac.cn).
M
I. INTRODUCTION
such experiments, there are always some problems. Firstly, the existence of noises in all datasets can not be avoided. There are two kinds of noises, i.e, systematic noise and random noise, with microarray data. Systematic noise derives from the experimental apparatus’ imperfection or non-calibration, ill-considered measurement, or the physical bent of the experimenters; Random noises are usually caused by casual factors. These noises affect the data greatly. Secondly, due to the small sample size problem in microarray dataset, a classification system may take risk of the over-fitting problem when used in a unique dataset. That is, the classifiers are trained on the dataset to an extreme level that they may achieve high classification accuracy only on the existing data, and fail to classify new test samples correctly, which results in poor generalization ability.
Based on these observations, it is of great value to train classifiers in one dataset and then validate them in others. In this way, both the generalization ability of classifiers and the biological significance of different genes can be evaluated objectively. In addition, the problems described above could be alleviated to some extent. Utility of datasets from different labs could do their best to circumvent the mistakes inherent to them. The more datasets the classifiers are trained from, the more common characteristics these classifiers will present. For the second problem mentioned above, the feature of microarray dataset that the amount of genes is much larger than that of samples makes it difficult to apply the existing methods to train classifiers. However, the combination of datasets could add the amount of samples to balance the uneven dimensions of microarray dataset. Third, there are a large amount of microarray datasets covering varieties of diseases in super datasets, such as EMBL, NCBI. It’s a big resource available to seek for disease genes. So, it’s feasible and essential to combine these datasets.
Some researchers have tried to merge datasets of different origins to yield generalized classifiers. In the paper by (Yu, et al, 2007) [5], classifiers are trained in one dataset and tested on the independent datasets from other labs. Because most of the informative genes used to build classifiers are excluded from the other datasets, few effective classifiers could be applied. And these applicable classifiers didn’t achieve good results.
In this paper, common genes and corresponding expression values of datasets are extracted first, and then utilized to train classifiers. The method proposed in this paper is based on genetic programming (GP). GP has been widely applied to deal with classification problems because it can discover underlying data relationships and present them mathematically. Among the evolutionary approaches, GP is a
978-1-4244-2756-7/09/$25.00 ©2009 IEEE
In this paper, we propose a genetic programming ...on the analysis of multiclass microarray datasets....using an evolutionary algorithm (EA) to select ...
In this paper, we investigate the feasibility of using Formal Concept Analysis (FCA) as a tool for microarray data analysis. The method of FCA builds ...
Genetic Programming based DNA Microarray Analysis for Classification of Cancer...The results indicate that using gene selection and GP as implemented in ...
Analysis of Microarray... 暂无评价 3页 1下载券 cudaBayesreg:Parallel... 暂无评价 24页 4下载券 Data association in mu... 17页 免费 Using Genetic ...
microarray data analysis is that the number of ...finding genetic markers that are due only to ...Programming for Link Discovery, Proceedings of the...
cation of cancer based on DNA microarray data that uses diverse ensemble ...cation rules obtained by genetic programming is useful to improve the ...
THE REPRESENTATION, RETRIEVAL, AND ANALYSIS OF MICROARRAY DATA TIMOTHY DANFORD...using our previously published “Joint Binding Deconvolution” (JBD) algorithm...
ProMAT Protein Microarray Analysis Tool_专业资料。...The models are fit using ‘nlm’, a nonlinear ..., an open-source statistical programming language....
2 DESCRIPTION 1 INTRODUCTION The analysis of large data sets has become a ... or can be created directly from microarray and other data using programs ...
Genetic Programming base... 暂无评价 14页 免费...using a variety of feature selection methods and ...The analysis of DNA microarray data is divided ...
热门文档