Gene and sample selection using T-score with sample selection

Gene selection from high-dimensional microarray gene-expression data is statistically a challenging problem. Filter approaches to gene selection have been popular because of their simplicity, efficiency, and accuracy. Due to small sample size, all samples are generally used to compute relevant ranking statistics and selection of samples in filter-based gene selection methods has not been addressed. In this paper, we extend previously-proposed simultaneous sample and gene selection approach. In a backward elimination method, a modified logistic regression loss function is used to select relevant samples at each iteration, and these samples are used to compute the T-score to rank genes. This method provides a compromise solution between T-score and other support vector machine (SVM) based algorithms. The performance is demonstrated on both simulated and real datasets with criteria such as classification performance, stability and redundancy. Results indicate that computational complexity and stability of the method are improved compared to SVM based methods without compromising the classification performance.     

Cancer classification and prediction using logistic regression with Bayesian gene selection

In microarray-based cancer classification and prediction, gene selection is an important research problem owing to the large number of genes and the small number of experimental conditions. In this paper, we propose a Bayesian approach to gene selection and classification using the logistic regression model. The basic idea of our approach is in conjunction with a logistic regression model to relate the gene expression with the class labels. We use Gibbs sampling and Markov chain Monte Carlo (MCMC) methods to discover important genes. To implement Gibbs Sampler and MCMC search, we derive a posterior distribution of selected genes given the observed data. After the important genes are identified, the same logistic regression model is then used for cancer classification and prediction. Issues for efficient implementation for the proposed method are discussed. The proposed method is evaluated against several large microarray data sets, including hereditary breast cancer, small round blue-cell tumors, and acute leukemia. The results show that the method can effectively identify important genes consistent with the known biological findings while the accuracy of the classification is also high. Finally, the robustness and sensitivity properties of the proposed method are also investigated.     

An efficient statistical feature selection approach for classification of gene expression data

Classification of gene expression data plays a significant role in prediction and diagnosis of diseases. Gene expression data has a special characteristic that there is a mismatch in gene dimension as opposed to sample dimension. All genes do not contribute for efficient classification of samples. A robust feature selection algorithm is required to identify the important genes which help in classifying the samples efficiently. In order to select informative genes (features) based on relevance and redundancy characteristics, many feature selection algorithms have been introduced in the past. Most of the earlier algorithms require computationally expensive search strategy to find an optimal feature subset. Existing feature selection methods are also sensitive to the evaluation measures. The paper introduces a novel and efficient feature selection approach based on statistically defined effective range of features for every class termed as ERGS (Effective Range based Gene Selection). The basic principle behind ERGS is that higher weight is given to the feature that discriminates the classes clearly. Experimental results on well-known gene expression datasets illustrate the effectiveness of the proposed approach. Two popular classifiers viz. Nave Bayes Classifier (NBC) and Support Vector Machine (SVM) have been used for classification. The proposed feature selection algorithm can be helpful in ranking the genes and also is capable of identifying the most relevant genes responsible for diseases like leukemia, colon tumor, lung cancer, diffuse large B-cell lymphoma (DLBCL), prostate cancer.     

An experimental comparison of gene selection by Lasso and Dantzig selector for cancer classification

Selecting a subset of genes with strong discriminative power is a very important step in classification problems based on gene expression data. Lasso and Dantzig selector are known to have automatic variable selection ability in linear regression analysis. This paper applies Lasso and Dantzig selector to select the most informative genes for representing the probability of an example being positive as a linear function of the gene expression data. The selected genes are further used to fit different classifiers for cancer classification. Comparative experiments were conducted on six publicly available cancer datasets, and the detailed comparison results show that in general, Lasso is more capable than Dantzig selector at selecting informative genes for cancer classification.     

Simultaneous genes and training samples selection by modified particle swarm optimization for gene expression data classification

Gene expression datasets is a means to classify and predict the diagnostic categories of a patient. Informative genes and representative samples selection are two important aspects for reducing gene expression data. Identifying and pruning redundant genes and samples simultaneously can improve the performance of classification and circumvent the local optima problem. In the present paper, the modified particle swarm optimization was applied to selecting optimal genes and samples simultaneously and support vector machine was used as an objective function to determine the optimum set of genes and samples. To evaluate the performance of the new proposed method, it was applied to three publicly available microarray datasets. It has been demonstrated that the proposed method for gene and sample selection is a useful tool for mining high dimension data.     

An integrated algorithm for gene selection and classification applied to microarray data of ovarian cancer

OBJECTIVE: The type of data in microarray provides unprecedented amount of data. A typical microarray data of ovarian cancer consists of the expressions of tens of thousands of genes on a genomic scale, and there is no systematic procedure to analyze this information instantaneously. To avoid higher computational complexity, it needs to select the most likely differentially expressed gene markers to explain the effects of ovarian cancer. Traditionally, gene markers are selected by ranking genes according to statistics or machine learning algorithms. In this paper, an integrated algorithm is derived for gene selection and classification in microarray data of ovarian cancer. METHODS: First, regression analysis is applied to find target genes. Genetic algorithm (GA), particle swarm optimization (PSO), support vector machine (SVM), and analysis of variance (ANOVA) are hybridized to select gene markers from target genes. Finally, the improved fuzzy model is applied to classify cancer tissues. RESULTS: The microarray data of ovarian cancer, obtained from China Medical University Hospital, is used to test the performance of the proposed algorithm. In simulation, 200 target genes are obtained after regression analysis and six gene markers are selected from the hybrid process of GA, PCO, SVM and ANOVA. Additionally, these gene markers are used to classify cancer tissues. CONCLUSIONS: The proposed algorithm can be used to analyze gene expressions and has superior performance in microarray data of ovarian cancer, and it can be performed on other studies for cancer diagnosis.     

Multiclass cancer classification by using fuzzy support vector machine and binary decision tree with gene selection

We investigate the problems of multiclass cancer classification with gene selection from gene expression data. Two different constructed multiclass classifiers with gene selection are proposed, which are fuzzy support vector machine (FSVM) with gene selection and binary classification tree based on SVM with gene selection. Using F test and recursive feature elimination based on SVM as gene selection methods, binary classification tree based on SVM with F test, binary classification tree based on SVM with recursive feature elimination based on SVM, and FSVM with recursive feature elimination based on SVM are tested in our experiments. To accelerate computation, preselecting the strongest genes is also used. The proposed techniques are applied to analyze breast cancer data, small round blue-cell tumors, and acute leukemia data. Compared to existing multiclass cancer classifiers and binary classification tree based on SVM with F test or binary classification tree based on SVM with recursive feature elimination based on SVM mentioned in this paper, FSVM based on recursive feature elimination based on SVM can find most important genes that affect certain types of cancer with high recognition accuracy.     

Multivariate selection of genetic markers in diagnostic classification

Analysis of gene expression data obtained from microarrays presents a new set of challenges to machine learning modeling. In this domain, in which the number of variables far exceeds the number of cases, identifying relevant genes or groups of genes that are good markers for a particular classification is as important as achieving good classification performance. Although several machine learning algorithms have been proposed to address the latter, identification of gene markers has not been systematically pursued. In this article, we investigate several algorithms for selecting gene markers for classification. We test these algorithms using logistic regression, as this is a simple and efficient supervised learning algorithm. We demonstrate, using 10 different data sets, that a conditionally univariate algorithm constitutes a viable choice if a researcher is interested in quickly determining a set of gene expression levels that can serve as markers for disease. We show that the classification performance of logistic regression is not very different from that of more sophisticated algorithms that have been applied in previous studies, and that the gene selection in the logistic regression algorithm is reasonable in both cases. Furthermore, the algorithm is simple, its theoretical basis is well established, and our user-friendly implementation is now freely available on the internet, serving as a benchmarking tool for the development of new algorithms.     

Simultaneous analysis of the gene expression profiles of cancer and stromal cells in endometrial cancer

To address the role of cancer‐stroma interactions, we performed gene expression profiling of both cancer and stroma, using matching samples of endometrial cancer (EC), and analyzed the relationship between the gene expression pattern and prognosis in EC. Sixty EC cases were included in this study (38 nonrecurrent and 22 recurrent). Cancer and stroma were separated by performing laser capture microdissection, and microarray analysis was performed separately on cancer and stromal cells. Genes related with progression‐free survival (PFS) in cancer and stroma were analyzed using the Cox regression model, and we established a formula, based on the gene expression pattern of cancer and stroma, to predict recurrence using logistic regression. We estimated the accuracy of the formula using the 0.632 method. All cases were classified based on the 79 selected genes of cancer and stroma related to PFS, based on unsupervised clustering. A total of 143 genes in cancer, and 79 genes in stroma were significantly related with PFS. The estimated area under the curve of receiver operating characteristics curve in cancer and stroma to predict recurrence were 0.800 and 0.758, respectively. Based on the 79 genes of cancer, the 22 recurrent cases were divided into two groups, which generally correlated with the histological grade. In contrast, based on the 79 genes of stroma, the 22 recurrent cases displayed homogeneous gene expression, unrelated to the histological grade. We conclude that gene expression profiles of cancer and stroma can predict the recurrence of EC and stromal that gene expression does not depend on the cancer grade. © 2014 Wiley Periodicals, Inc.     

Gene expression data classification with Kernel principal component analysis

One important feature of the gene expression data is that the number of genes M far exceeds the number of samples N. Standard statistical methods do not work well when N < M. Development of new methodologies or modification of existing methodologies is needed for the analysis of the microarray data. In this paper, we propose a novel analysis procedure for classifying the gene expression data. This procedure involves dimension reduction using kernel principal component analysis (KPCA) and classification with logistic regression (discrimination). KPCA is a generalization and nonlinear version of principal component analysis. The proposed algorithm was applied to five different gene expression datasets involving human tumor samples. Comparison with other popular classification methods such as support vector machines and neural networks shows that our algorithm is very promising in classifying gene expression data.     

Ensemble gene selection by grouping for microarray data classification

Selecting relevant and discriminative genes for sample classification is a common and critical task in gene expression analysis (e.g. disease diagnostic). It is desirable that gene selection can improve classification performance of learning algorithm effectively. In general, for most gene selection methods widely used in reality, an individual gene subset will be chosen according to its discriminative power. One of deficiencies of individual gene subset is that its contribution to classification purpose is limited. This issue can be alleviated by ensemble gene selection based on random selection to some extend. However, the random one requires an unnecessary large number of candidate gene subsets and its reliability is a problem. In this study, we propose a new ensemble method, called ensemble gene selection by grouping (EGSG), to select multiple gene subsets for the classification purpose. Rather than selecting randomly, our method chooses salient gene subsets from microarray data by virtue of information theory and approximate Markov blanket. The effectiveness and accuracy of our method is validated by experiments on five publicly available microarray data sets. The experimental results show that our ensemble gene selection method has comparable classification performance to other gene selection methods, and is more stable than the random one.     

Selecting significant genes by randomization test for cancer classification using gene expression data

Gene selection is an important task in bioinformatics studies, because the accuracy of cancer classification generally depends upon the genes that have biological relevance to the classifying problems. In this work, randomization test (RT) is used as a gene selection method for dealing with gene expression data. In the method, a statistic derived from the statistics of the regression coefficients in a series of partial least squares discriminant analysis (PLSDA) models is used to evaluate the significance of the genes. Informative genes are selected for classifying the four gene expression datasets of prostate cancer, lung cancer, leukemia and non-small cell lung cancer (NSCLC) and the rationality of the results is validated by multiple linear regression (MLR) modeling and principal component analysis (PCA). With the selected genes, satisfactory results can be obtained.     

Computational selection of distinct class- and subclass-specific gene expression signatures

In this investigation we used statistical methods to select genes with expression profiles that partition classes and subclasses of biological samples. Gene expression data corresponding to liver samples from rats treated for 24 h with an enzyme inducer (phenobarbital) or a peroxisome proliferator (clofibrate, gemfibrozil or Wyeth 14,643) were subjected to a modified Z-score test to identify gene outliers and a binomial distribution to reduce the probability of detecting genes as differentially expressed by chance. Hierarchical clustering of 238 statistically valid differentially expressed genes partitioned class-specific gene expression signatures into groups that clustered samples exposed to the enzyme inducer or to peroxisome proliferators. Using analysis of variance (ANOVA) and linear discriminant analysis methods we identified single genes as well as coupled gene expression profiles that separated the phenobarbital from the peroxisome proliferator treated samples and discerned the fibrate (gemfibrozil and clofibrate) subclass of peroxisome proliferators. A comparison of genes ranked by ANOVA with genes assessed as significant by mixedlinear models analysis [J. Comput. Biol. 8 (2001) 625] or ranked by information gain revealed good congruence with the top 10 genes from each statistical method in the contrast between phenobarbital and peroxisome proliferators expression profiles. We propose building upon a classification regimen comprised of analysis of replicate data, outlier diagnostics and gene selection procedures to utilize cDNA microarray data to categorize subclasses of samples exposed to pharmacologic agents.     

A simulation to analyze feature selection methods utilizing gene ontology for gene expression classification

Gene expression profile classification is a pivotal research domain assisting in the transformation from traditional to personalized medicine. A major challenge associated with gene expression data classification is the small number of samples relative to the large number of genes. To address this problem, researchers have devised various feature selection algorithms to reduce the number of genes. Recent studies have been experimenting with the use of semantic similarity between genes in Gene Ontology (GO) as a method to improve feature selection. While there are few studies that discuss how to use GO for feature selection, there is no simulation study that addresses when to use GO-based feature selection. To investigate this, we developed a novel simulation, which generates binary class datasets, where the differentially expressed genes between two classes have some underlying relationship in GO. This allows us to investigate the effects of various factors such as the relative connectedness of the underlying genes in GO, the mean magnitude of separation between differentially expressed genes denoted by δ, and the number of training samples. Our simulation results suggest that the connectedness in GO of the differentially expressed genes for a biological condition is the primary factor for determining the efficacy of GO-based feature selection. In particular, as the connectedness of differentially expressed genes increases, the classification accuracy improvement increases. To quantify this notion of connectedness, we defined a measure called Biological Condition Annotation Level BCAL(G), where G is a graph of differentially expressed genes. Our main conclusions with respect to GO-based feature selection are the following: (1) it increases classification accuracy when BCAL(G) ⩾ 0.696; (2) it decreases classification accuracy when BCAL(G) ⩽ 0.389; (3) it provides marginal accuracy improvement when 0.389 < BCAL(G) < 0.696 and δ < 1; (4) as the number of genes in a biological condition increases beyond 50 and δ ⩾ 0.7, the improvement from GO-based feature selection decreases; and (5) we recommend not using GO-based feature selection when a biological condition has less than ten genes. Our results are derived from datasets preprocessed using RMA (Robust Multi-array Average), cases where δ is between 0.3 and 2.5, and training sample sizes between 20 and 200, therefore our conclusions are limited to these specifications. Overall, this simulation is innovative and addresses the question of when SoFoCles-style feature selection should be used for classification instead of statistical-based ranking measures.     

Interactive exploration of microarray gene expression patterns in a reduced dimensional space

The very high dimensional space of gene expression measurements obtained by DNA microarrays impedes the detection of underlying patterns in gene expression data and the identification of discriminatory genes. In this paper we show the use of projection methods such as principal components analysis (PCA) to obtain a direct link between patterns in the genes and patterns in samples. This feature is useful in the initial interactive pattern exploration of gene expression data and data-driven learning of the nature and types of samples. Using oligonucleotide microarray measurements of 40 samples from different normal human tissues, we show that distinct patterns are obtained when the genes are projected on a two-dimensional plane spanned by the loadings of the two major principal components. These patterns define the particular genes associated with a sample class (i.e., tissue). When used separately from the other genes, these class-specific (i.e., tissue-specific) genes in turn define distinct tissue patterns in the projection space spanned by the scores of the two major principal components. In this study, PCA projection facilitated discriminatory gene selection for different tissues and identified tissue-specific gene expression signatures for liver, skeletal muscle, and brain samples. Furthermore, it allowed the classification of nine new samples belonging to these three types using the linear combination of the expression levels of the tissue-specific genes determined from the first set of samples. The application of the technique to other published data sets is also discussed.     

A kernel-based clustering method for gene selection with gene expression data

Gene selection is important for cancer classification based on gene expression data, because of high dimensionality and small sample size. In this paper, we present a new gene selection method based on clustering, in which dissimilarity measures are obtained through kernel functions. It searches for best weights of genes iteratively at the same time to optimize the clustering objective function. Adaptive distance is used in the process, which is suitable to learn the weights of genes during the clustering process, improving the performance of the algorithm. The proposed algorithm is simple and does not require any modification or parameter optimization for each dataset. We tested it on eight publicly available datasets, using two classifiers (support vector machine, k-nearest neighbor), compared with other six competitive feature selectors. The results show that the proposed algorithm is capable of achieving better accuracies and may be an efficient tool for finding possible biomarkers from gene expression data.     

A new classification model with simple decision rule for discovering optimal feature gene pairs

Classifiers have been widely used to select an optimal subset of feature genes from microarray data for accurate classification of cancer samples and cancer-related studies. However, the classification rules derived from most classifiers are complex and difficult to understand in biological significance. How to solve this problem is a new challenge. In this paper, a new classification model based on gene pair is proposed to address the problem. The experimental results on several microarray data demonstrate that the proposed classification model performs well in finding a large number of excellent feature gene pairs. A 100% LOOCV classification accuracy can be achieved using a single classification model based on optimal feature gene pair or combining multiple top-ranked classification models. Using the proposed method, we successfully identified important cancer-related genes that had been validated in previous biological studies while they were not discovered by the other methods.     

基于粒子群优化算法的白血病基因表达样本分类研究

目的基于分子生物学的微阵列基因表达数据和智能优化算法对白血病肿瘤样本进行分类研究。方法给出基于粒子群优化(PSO)算法用于分类模型的训练和测试,选取含7129个基因的72个白血病基因表达样本,从中选取包含50、100和200个特征基因的3组数据,在不同基因数条件下分别执行10次分类测试。建立基于K-均值算法的分类模型,在同等条件下验证PSO算法分类性能。使用准确率、精确率、召回率、F1值等机器学习指标及Boxplot和Heatmap图谱用于分析对比。结果PSO算法用于分类测试的数据分别含20例急性淋巴细胞白血病(ALL)和14例急性髓细胞白血病(AML)样本。10次分类结果的平均分类准确率均在90%左右;PSO算法的分类准确率并不稳定,10次分类测试中,准确率的平均值和最优值间存在明显差异;ALL亚型的召回率明显高于AML亚型,均接近100%,但AML亚型的精确率明显高于ALL亚型,均接近100%,F1值可比性不大。K-均值算法与PSO算法类似,分类性能随着基因数的增加而降低;K-均值算法在200基因数条件下分类结果较差,分类稳定性和准确率均出现大幅下降,且低于同等条件下PSO算法分类结果;100个基因数条件下,ALL亚型召回率为100%,高于AML亚型;AML亚型精确率为100%,高于ALL亚型;200个基因数条件下,平均值中ALL亚型召回率和F1值高于AML亚型,AML亚型精确率高于ALL亚型,其最优值的统计学指标差异不大。相同白血病肿瘤样本的不同特征基因数条件下,PSO算法可获得较高准确率的分类结果,但分类稳定性不足,整体上优于K-均值算法。结论PSO算法能够应用于白血病基因表达样本的分类研究。