A mixture model with a reference-based automatic selection of components for disease classification from protein and/or gene expression levels

Abstract

BackgroundBioinformatics data analysis is often using linear mixture model representing samples as additive mixture of components. Properly constrained blind matrix factorization methods extract those components using mixture samples only. However, automatic selection of extracted components to be retained for classification analysis remains an open issue.ResultsThe method proposed here is applied to well-studied protein and genomic datasets of ovarian, prostate and colon cancers to extract components for disease prediction. It achieves average sensitivities of: 96.2 (sd = 2.7%), 97.6% (sd = 2.8%) and 90.8% (sd = 5.5%) and average specificities of: 93.6% (sd = 4.1%), 99% (sd = 2.2%) and 79.4% (sd = 9.8%) in 100 independent two-fold cross-validations.ConclusionsWe propose an additive mixture model of a sample for feature extraction using, in principle, sparseness constrained factorization on a sample-by-sample basis. As opposed to that, existing methods factorize complete dataset simultaneously. The sample model is composed of a reference sample representing control and/or case (disease) groups and a test sample. Each sample is decomposed into two or more components that are selected automatically (without using label information) as control specific, case specific and not differentially expressed (neutral). The number of components is determined by cross-validation. Automatic assignment of features (m/z ratios or genes) to particular component is based on thresholds estimated from each sample directly. Due to the locality of decomposition, the strength of the expression of each feature across the samples can vary. Yet, they will still be allocated to the related disease and/or control specific component. Since label information is not used in the selection process, case and control specific components can be used for classification. That is not the case with standard factorization methods. Moreover, the component selected by proposed method as disease specific can be interpreted as a sub-mode and retained for further analysis to identify potential biomarkers. As opposed to standard matrix factorization methods this can be achieved on a sample (experiment)-by-sample basis. Postulating one or more components with indifferent features enables their removal from disease and control specific components on a sample-by-sample basis. This yields selected components with reduced complexity and generally, it increases prediction accuracy.

A Nonlinear Mixture Model based Unsupervised Variable Selection in Genomics and Proteomics

“…The mixture model is comprised of considered sample and a reference sample that represents negative (healthy) class. The model is nonlinear generalization of the linear mixture model with a reference sample that was presented in (Kopriva and Filipović, 2011).…”

“…Proposed method yields comparable accuracy with slightly more variables than supervised methods and it outperforms its linear counterpart (Kopriva and Filipović, 2011).…”

“…they are sparse (Schachtner et al, 2008;Stadtlthanner et al, 2008;Gao and Church, 2005;Kim and Park, 2007). NMF (Cichocki et al, 2010) decomposes dataset into nonnegative components (Brunet et al, 2004) that in addition can be sparse (Gao and Church, 2005;Kim and Park, 2007;Kopriva and Filipović, 2011). Unsupervised decomposition methods for variable selection, the representatives of which are referenced above, have the following limitations: (i) they are based on a linear mixture model (Girolami and Breitlling, 2004) representing dataset as weighted linear superposition of components.…”

“…Thereby, decomposition is based on nonlinear mixture model comprised of considered sample and a reference sample representing negative (healthy) class. The model is nonlinear generalization of the linear mixture model with a reference sample presented in (Kopriva and Filipović, 2011). Nonlinear mapping is performed across sample dimension yielding possibly linear model with preserved number of variables and "increased" number of samples.…”

“…Hence, it is believed that performance estimate of proposed approach to variable selection is trustworthy. Proposed approach is compared with three state-of-the-art supervised (Brown 2009;Aliferis et al, 2010) and one unsupervised (Kopriva and Filipović, 2011) variable selection method on three well-known cancer types in genomics: colon cancer (Alon et al, 1999), diffuse large b-cell lymphomas and follicular lymphomas (Shipp et al, 2002) and prostate cancer (Singh et al, 2002), and two well-known cancer types in proteomics: ovarian cancer (Petricoin et al, 2002a) and prostate cancer (Petricoin et al, 2002b).…”