MOTIVATION: Due to the high complexity of metabolome, the comprehensive 2D gas chromatography time-of-flight mass spectrometry (GC×GC-TOF MS) is considered as a powerful analytical platform for metabolomics study. However, the applications of GC×GC-TOF MS in metabolomics are not popular owing to the lack of bioinformatics system for data analysis. RESULTS: We developed a computational platform entitled metabolomics profiling pipeline (MetPP) for analysis of metabolomics data acquired on a GC×GC-TOF MS system. MetPP can process peak filtering and merging, retention index matching, peak list alignment, normalization, statistical significance tests and pattern recognition, using the peak lists deconvoluted from the instrument data as its input. The performance of MetPP software was tested with two sets of experimental data acquired in a spike-in experiment and a biomarker discovery experiment, respectively. MetPP not only correctly aligned the spiked-in metabolite standards from the experimental data, but also correctly recognized their concentration difference between sample groups. For analysis of the biomarker discovery data, 15 metabolites were recognized with significant concentration difference between the sample groups and these results agree with the literature results of histological analysis, demonstrating the effectiveness of applying MetPP software for disease biomarker discovery. AVAILABILITY: The source code of MetPP is available at http://metaopen.sourceforge.net CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: The compound identification in gas chromatography-mass spectrometry (GC-MS) is achieved by matching the experimental mass spectrum to the mass spectra in a spectral library. It is known that the intensities with higher m/z value in the GC-MS mass spectrum are the most diagnostic. Therefore, to increase the relative significance of peak intensities of higher m/z value, the intensities and m/z values are usually transformed with a set of weight factors. A poor quality of weight factors can significantly decrease the accuracy of compound identification. With the significant enrichment of the mass spectral database and the broad application of GC-MS, it is important to re-visit the methods of discovering the optimal weight factors for high confident compound identification. RESULTS: We developed a novel approach to finding the optimal weight factors only through a reference library for high accuracy compound identification. The developed approach first calculates the ratio of skewness to kurtosis of the mass spectral similarity scores among spectra (compounds) in a reference library and then considers a weight factor with the maximum ratio as the optimal weight factor. We examined our approach by comparing the accuracy of compound identification using the mass spectral library maintained by the National Institute of Standards and Technology. The results demonstrate that the optimal weight factors for fragment ion peak intensity and m/z value found by the developed approach outperform the current weight factors for compound identification. AVAILABILITY: The results and R package are available at http://stage.louisville.edu/faculty/x0zhan17/software/ software-development.

BACKGROUND: Comprehensive two-dimensional gas chromatography coupled with mass spectrometry (GC × GC-MS) is a powerful technique which has gained increasing attention over the last two decades. The GC × GC-MS provides much increased separation capacity, chemical selectivity and sensitivity for complex sample analysis and brings more accurate information about compound retention times and mass spectra. Despite these advantages, the retention times of the resolved peaks on the two-dimensional gas chromatographic columns are always shifted due to experimental variations, introducing difficulty in the data processing for metabolomics analysis. Therefore, the retention time variation must be adjusted in order to compare multiple metabolic profiles obtained from different conditions. RESULTS: We developed novel peak alignment algorithms for both homogeneous (acquired under the identical experimental conditions) and heterogeneous (acquired under the different experimental conditions) GC × GC-MS data using modified Smith-Waterman local alignment algorithms along with mass spectral similarity. Compared with literature reported algorithms, the proposed algorithms eliminated the detection of landmark peaks and the usage of retention time transformation. Furthermore, an automated peak alignment software package was established by implementing a likelihood function for optimal peak alignment. CONCLUSIONS: The proposed Smith-Waterman local alignment-based algorithms are capable of aligning both the homogeneous and heterogeneous data of multiple GC × GC-MS experiments without the transformation of retention times and the selection of landmark peaks. An optimal version of the SW-based algorithms was also established based on the associated likelihood function for the automatic peak alignment. The proposed alignment algorithms outperform the literature reported alignment method by analyzing the experiment data of a mixture of compound standards and a metabolite extract of mouse plasma with spiked-in compound standards.

BACKGROUND: Diagnosis techniques using urine are non-invasive, inexpensive, and easy to perform in clinical settings. The metabolites in urine, as the end products of cellular processes, are closely linked to phenotypes. Therefore, urine metabolome is very useful in marker discoveries and clinical applications. However, only univariate methods have been used in classification studies using urine metabolome. Since multiple genes or proteins would be involved in developments of complex diseases such as breast cancer, multiple compounds including metabolites would be related with the complex diseases, and multivariate methods would be needed to identify those multiple metabolite markers. Moreover, because combinatorial effects among the markers can seriously affect disease developments and there also exist individual differences in genetic makeup or heterogeneity in cancer progressions, single marker is not enough to identify cancers. RESULTS: We proposed classification models using multivariate classification techniques and developed an analysis procedure for classification studies using metabolome data. Through this strategy, we identified five potential urinary biomarkers for breast cancer with high accuracy, among which the four biomarker candidates were not identifiable by only univariate methods. We also proposed potential diagnosis rules to help in clinical decision making. Besides, we showed that combinatorial effects among multiple biomarkers can enhance discriminative power for breast cancer. CONCLUSIONS: In this study, we successfully showed that multivariate classifications are needed to precisely diagnose breast cancer. After further validation with independent cohorts and experimental confirmation, these marker candidates will likely lead to clinically applicable assays for earlier diagnoses of breast cancer.

The prediction risk estimation in nonlinear regres- sion models including artificial neural networks is especially important for problems with limited data since it can be used as a tool for finding the optimal model (or network architecture) minimizing the expected risk. In this paper, we suggest the prediction risk bounds of nonlinear regression models. The suggested bounds are derived from the modulus of continuity for a multivariate function. We also present the model selection criteria referred to as the modulus of continuity information criteria (MCIC) derived from the suggested prediction risk bounds. Through the simulation for function approximation, we have shown that the suggested MCIC is effective in nonlinear model selection problems with limited data.

This paper presents a new form of true risk bounds for the regression of real-valued functions. The goal of machine learning is minimizing the true risk (or general error) for the whole distribution of sample space, not just a set of training samples. However, the true risk cannot be estimated accurately with the finite number of samples. In this sense, we derive the form of true risk bounds which may provide the useful guideline for the optimization of learning models. Through the simulation for the function approximation, we have shown that the prediction of true risk bounds based on the suggested functional norm is well fitted to the empirical data.