Search

Abstract: Interpretation and mining of complex metabonomic data depend heavily on proper use of chemometric methods. Due to the “small n” paradigm and the absence of sufficient information concerning distribution of data, the classical parametric methods based on known theoretical distributions are sometimes unsuitable or unreliable to treat such data. Therefore, nonparametric methods requiring no or very limited assumptions provide useful alternative tools in many practical applications. In this paper, a new discriminant partial least squares combined with bootstrap and uninformative variable elimination (DPLS–BS–UVE) method is proposed for biomarker discovery in metabonomics. The method was tested on two real chromatographic data sets containing plasma metabolic profilings for S180 and H22 tumor-bearing mice. A robust version of c j was used as the cutoff criterion. The results of biomarker discovery were compared with those obtained using variable importance in the projection (VIP) as well as BS. It is demonstrated that similar results are obtained using the three methods and DPLS–BS–UVE could provide easy interpretation of raw data. When the resampling unit increases to 500, the results were not significantly affected. In conclusion, DPLS–BS–UVE is a reliable alternative method for biomarker discovery, especially when the sample size is small. [Copyright &y& Elsevier]

Biomarker selection has played an increasingly important part in modern medicine with advances of omics techniques. Kohonen self-organizing map is a well-established variable reduction algorithm in identifying significant biomarkers based on variable clustering. However, high dimensionality but small sample size of omics data makes self-organizing map-based model problematic in terms of selection stability and reproducibility. A novel feature screening system is presented in this study by coupling bootstrap with synergy self-organizing map-based orthogonal partial least squares discriminant analysis for stable and biologically meaningful metabolic biomarker selection. In the proposed feature screening system, particle swarm optimization algorithm is utilized to configure synergy self-organizing map-based orthogonal partial least squares discriminant analysis to perform the combination of clusters in a heuristic learning manner, enabling flexible selection of more informative features cost-effectively. Based on the paradigm of ensemble feature selection, bootstrap is adopted to explore significant variables consistently identified across multiple feature selectors rather than a single one. The feasibility of the novel feature screening system is evaluated by two most common inherited metabolic diseases, methylmalonic academia and propionic academia, using urinary metabolomics data. With the desirable classification performance, the proposed feature screening system outperforms simpler techniques in the identification of more features closely correlated with the metabolic mechanisms and the stability of selected candidate biomarkers against sample variations. Besides, the novel feature screening system greatly degrades the sensitivity of identified candidate biomarkers to the network size of self-organizing map, benefiting the identification of a suitable and stable final candidate biomarker list. Image 1 • A novel feature screening system was developed for stable metabolic biomarker selection. • Strategies of cluster combination and ensemble feature selection were involved. • Urinary metabolomics data of methylmalonic academia and propionic academia samples were modeled. • Desirable results were obtained in stable metabolic biomarker selection for inherited metabolic diseases. [ABSTRACT FROM AUTHOR]