Approaches to dimensionality reduction in proteomic biomarker studies

Abstract

Mass-spectra based proteomic profiles have received widespread attention as potential tools for biomarker discovery and early disease diagnosis. A major data-analytical problem involved is the extremely high dimensionality (i.e. number of features or variables) of proteomic data, in particular when the sample size is small. This article reviews dimensionality reduction methods that have been used in proteomic biomarker studies. It then focuses on the problem of selecting the most appropriate method for a specific task or dataset, and proposes method combination as a potential alternative to single-method selection. Finally, it points out the potential of novel dimension reduction techniques, in particular those that incorporate domain knowledge through the use of informative priors or causal inference.

Stable feature selection for biomarker discovery

“…discriminating healthy versus diseased, or different tumor stages) [3,4]. In the language of statistics and machine learning, this is often referred to as feature selection.…”

“…Feature selection has attracted strong research interest in the past several decades. For recent reviews of feature selection techniques used in bioinformatics, the reader is referred to [5,6,3,7].…”

“…• Compared to current feature selection reviews [5,6,3,7], this review focuses only on those feature selection approaches that incorporate "stability" into the algorithmic design.…”

Large margin classification with indefinite similarities

“…It was rediscovered under many guises later, such as for being a special case of the generalized SVM (Mangasarian 1998) and for being a method of embedding similarities into features (Chen et al 2009a;Pekalska et al 2001). However, it is widely used in the literature for embedded feature selection (Bradley and Mangasarian 1998;Zhu et al 2004;Fung and Mangasarian 2004;Zou 2007;Hilario and Kalousis 2008;Liu et al 2010).…”

“…In the literature, 1-norm SVM is often used as an embedded feature selection method, where learning and feature selection are performed simultaneously (Bradley and Mangasarian 1998;Zhu et al 2004;Fung and Mangasarian 2004;Zou 2007;Hilario and Kalousis 2008;Liu et al 2010). It was studied in Zhu et al (2004), where it was argued that 1-norm SVM has an advantage over the standard form of SVM in (1) when there are redundant noisy features.…”