A comparative study of feature selection and multiclass classification methods for tissue classification based on gene expression

Abstract

This paper studies the problem of building multiclass classifiers for tissue classification based on gene expression. The recent development of microarray technologies has enabled biologists to quantify gene expression of tens of thousands of genes in a single experiment. Biologists have begun collecting gene expression for a large number of samples. One of the urgent issues in the use of microarray data is to develop methods for characterizing samples based on their gene expression. The most basic step in the research direction is binary sample classification, which has been studied extensively over the past few years. This paper investigates the next step-multiclass classification of samples based on gene expression. The characteristics of expression data (e.g. large number of genes with small sample size) makes the classification problem more challenging. The process of building multiclass classifiers is divided into two components: (i) selection of the features (i.e. genes) to be used for training and testing and (ii) selection of the classification method. This paper compares various feature selection methods as well as various state-of-the-art classification methods on various multiclass gene expression datasets. Our study indicates that multiclass classification problem is much more difficult than the binary one for the gene expression datasets. The difficulty lies in the fact that the data are of high dimensionality and that the sample size is small. The classification accuracy appears to degrade very rapidly as the number of classes increases. In particular, the accuracy was very low regardless of the choices of the methods for large-class datasets (e.g. NCI60 and GCM). While increasing the number of samples is a plausible solution to the problem of accuracy degradation, it is important to develop algorithms that are able to analyze effectively multiple-class expression data for these special datasets.

Detecting reliable gene interactions by a hierarchy of Bayesian network classifiers

“…For instance, in the Leukemia set results, at a confidence level of t = 200, four variables with four arcs correctly predict 70% of the samples; in the CRC domain, at a level of t = 310, the estimation of the accuracy with only four variables and three arcs achieves a mean value of 96%. This fact corroborates other studies regarding gene expression classification based on a reduced number of genes [4,12,48]. The cardinality of the highest configured arc is included.…”

Data Visualization with Simultaneous Feature Selection

“…Initialize the mixing coefficient, α m , for each component, m, in the grid to 1/M ; Set the mean and the variance of the shared distribution, q(·|λ), as the mean and covariance of the training set; repeat Compute R, U and V using (6), (7) and (8) respectively, using current parameters, Θ; (9); end Obtain the center, µ m , of each component, m, of the mixture in the data space, using (11); Reestimate the width of the diagonal Gaussians, σ d , using (12), for all the features; Reestimate the mean and the variance of the shared distribution using (13) and (14) respectively; Reestimate the feature weight, ρ d , using (15), for all the features; until convergence; end The parameters are estimated using a variant of the EM algorithm as follows.…”

Variable selection methods for multi-class classification using signomial function

“…Filter methods score the merits of variables using intrinsic data properties such as information, distance, dependency and consistency, and then select a subset of variables as a preprocessing step independently of the choice of learning machine (Dhillon, et al, 2003;Torkkola, 2003;Li, et al, 2004;Yang and Pedersen 1997;Bolon-Canedo et al, 2012;Forman, 2004;You and Li, 2011;Rajapakse and Mundra, 2013). Filter methods usually are fast, but because they do not consider variable subsets' effects on the learning process, they can select a redundant one.…”