Background Gene selection is an important step when building predictors of disease state based on gene expression data. Gene selection generally improves performance and identifies a relevant subset of genes. Many univariate and multivariate gene selection approaches have been proposed. Frequently the claim is made that genes are co-regulated (due to pathway dependencies) and that multivariate approaches are therefore per definition more desirable than univariate selection approaches. Based on the published performances of all these approaches a fair comparison of the available results can not be made. This mainly stems from two factors. First, the results are often biased, since the validation set is in one way or another involved in training the predictor, resulting in optimistically biased performance estimates. Second, the published results are often based on a small number of relatively simple datasets. Consequently no generally applicable conclusions can be drawn. Results In this study we adopted an unbiased protocol to perform a fair comparison of frequently used multivariate and univariate gene selection techniques, in combination with a ränge of classifiers. Our conclusions are based on seven gene expression datasets, across several cancer types. Conclusion Our experiments illustrate that, contrary to several previous studies, in five of the seven datasets univariate selection approaches yield consistently better results than multivariate approaches. The simplest multivariate selection approach, the Top Scoring method, achieves the best results on the remaining two datasets. We conclude that the correlation structures, if present, are difficult to extract due to the small number of samples, and that consequently, overly-complex gene selection algorithms that attempt to extract these structures are prone to overtraining.@en

In a growing number of domains data captured encapsulates as many features as possible. This poses a challenge to classical pattern recognition techniques, since the number of samples often still is limited with respect to the number of features. Classical pattern recognition methods suffer from the small sample size, and robust classification techniques are needed. In order to reduce the dimensionality of the feature space, the selection of informative features becomes an essential step towards the classification task. The relevance of the features can be evaluated either individually (univariate approaches), or in a multivariate manner. Univariate approaches are simple and fast, therefore appealing. However, possible correlation and dependencies between the features are not considered. Therefore, multivariate search techniques may be helpful. Several limitations restrict the use of multivariate searches. First, they are prone to overtraining, especially in p n (many features and few samples) settings. Secondly, they can be computationally too expensive when dealing with a large feature space. We introduce a new multivariate search technique, that is less sensitive to the noise in the data and computationally feasible as well. We compare our approach with several multivariate and univariate feature selection techniques, on an artificial dataset which provides us with ground truth information, and on a real dataset. The results show the importance of multivariate search techniques and the robustness and reliability of our novel multivariate feature selection method.@en