Analysis of data from multiple sources has the potential to enhance knowledge discovery by capturing underlying structures, which are, otherwise, difficult to extract. Fusing data from multiple sources has already proved useful in many applications in social network analysis, signal processing and bioinformatics. However, data fusion is challenging since data from multiple sources are often (i) heterogeneous (i.e., in the form of higher-order tensors and matrices), (ii) incomplete, and (iii) have both shared and unshared components. In order to address these challenges, in this paper, we introduce a novel unsupervised data fusion model based on joint factorization of matrices and higher-order tensors. While the traditional formulation of coupled matrix and tensor factorizations modeling only shared factors fails to capture the underlying structures in the presence of both shared and unshared factors, the proposed data fusion model has the potential to automatically reveal shared and unshared components through modeling constraints. Using numerical experiments, we demonstrate the effectiveness of the proposed approach in terms of identifying shared and unshared components. Furthermore, we measure a set of mixtures with known chemical composition using both LC-MS (Liquid Chromatography - Mass Spectrometry) and NMR (Nuclear Magnetic Resonance) and demonstrate that the structure-revealing data fusion model can (i) successfully capture the chemicals in the mixtures and extract the relative concentrations of the chemicals accurately, (ii) provide promising results in terms of identifying shared and unshared chemicals, and (iii) reveal the relevant patterns in LC-MS by coupling with the diffusion NMR data. We have proposed a structure-revealing data fusion model that can jointly analyze heterogeneous, incomplete data sets with shared and unshared components and demonstrated its promising performance as well as potential limitations on both simulated and real data.
The metabolic composition of plasma is affected by time passed since the last meal and by individual variation in metabolite clearance rates. Rat plasma in fed and fasted states was analyzed with liquid chromatography quadrupole-time-of-flight mass spectrometry (LC-QTOF) for an untargeted investigation of these metabolite patterns. The dataset was used to investigate the effect of data preprocessing on biomarker selection using three different softwares, MarkerLynxTM, MZmine, XCMS along with a customized preprocessing method that performs binning of m/z channels followed by summation through retention time. Direct comparison of selected features representing the fed or fasted state showed large differences between the softwares. Many false positive markers were obtained from custom data preprocessing compared with dedicated softwares while MarkerLynxTM provided better coverage of markers. However, marker selection was more reliable with the gap filling (or peak finding) algorithms present in MZmine and XCMS. Further identification of the putative markers revealed that many of the differences between the markers selected were due to variations in features representing adducts or daughter ions of the same metabolites or of compounds from the same chemical subclasses, e.g., lyso-phosphatidylcholines (LPCs) and lyso-phosphatidylethanolamines (LPEs). We conclude that despite considerable differences in the performance of the preprocessing tools we could extract the same biological information by any of them. Carnitine, branched-chain amino acids, LPCs and LPEs were identified by all methods as markers of the fed state whereas acetylcarnitine was abundant during fasting in rats.