Automated Recommendation of Research Keywords from PubMed That Suggest the Molecular Mechanism Associated with Biomarker Metabolites.

Metabolomics can help identify candidate biomarker metabolites whose levels are altered in response to disease development or drug administration. However, assessment of the underlying molecular mechanism is challenging considering it depends on the researcher's knowledge. This study reports a novel method for the automated recommendation of keywords known in the literature that may be overlooked by researchers. The proposed method aided in the identification of Medical Subject Headings (MeSH) terms in PubMed using MeSH co-occurrence data. The intended users are biocurators who have identified specific biomarker metabolites from a metabolomics study and would like to identify literature-reported molecular mechanisms that are associated with both the metabolite and their research area of interest. The proposed method finds MeSH terms that co-occur with a MeSH term of the candidate biomarker metabolite as well as a MeSH term of a researcher's known keyword, such as the name of a disease. The connectivity score S was determined using association analysis. Pilot analyses demonstrated that, while the biological significance of the obtained MeSH terms could not be guaranteed, the developed method can be useful for finding keywords to further investigate molecular mechanisms in association with candidate biomarker molecules.

Fake metabolomics chromatogram generation for facilitating deep learning of peak-picking neural networks.

Finding peaks in chromatograms and determining their start and end points (peak picking) is a core task in chromatography based biotechnology. Construction of peak-picking neural networks by deep learning was, however, hampered from the preparation of exact peak-picked or "labeled" chromatograms since the exact start and end points were often unclear in overlapping peaks in real chromatograms. We present a design of a fake chromatogram generator, along with a method for deep learning of peak-picking neural networks. Fake chromatograms were generated by generation of fake peaks, random sampling of peak positions from feature distributions, and merging with real blank sample chromatograms. Information on the exact start and end points, as labeled on the fake chromatograms, were effective for training and evaluating peak-picking neural networks. The peak-picking neural networks constructed herein outperformed conventional peak-picking software and showed comparable performance with that of experienced operators for processing the widely targeted metabolome data. Results of this study indicate that generation of fake chromatograms would be crucial for developing peak-picking neural networks and a key technology for further improvement of peak picking neural networks.

Machine-learning Approach for the Development of a Novel Predictive Model for the Diagnosis of Hepatocellular Carcinoma.

Because of its multifactorial nature, predicting the presence of cancer using a single biomarker is difficult. We aimed to establish a novel machine-learning model for predicting hepatocellular carcinoma (HCC) using real-world data obtained during clinical practice. To establish a predictive model, we developed a machine-learning framework which developed optimized classifiers and their respective hyperparameter, depending on the nature of the data, using a grid-search method. We applied the current framework to 539 and 1043 patients with and without HCC to develop a predictive model for the diagnosis of HCC. Using the optimal hyperparameter, gradient boosting provided the highest predictive accuracy for the presence of HCC (87.34%) and produced an area under the curve (AUC) of 0.940. Using cut-offs of 200 ng/mL for AFP, 40 mAu/mL for DCP, and 15% for AFP-L3, the accuracies of AFP, DCP, and AFP-L3 for predicting HCC were 70.67% (AUC, 0.766), 74.91% (AUC, 0.644), and 71.05% (AUC, 0.683), respectively. A novel predictive model using a machine-learning approach reduced the misclassification rate by about half compared with a single tumor marker. The framework used in the current study can be applied to various kinds of data, thus potentially become a translational mechanism between academic research and clinical practice.

Serum immunoglobulin G Fc region N-glycosylation profiling by matrix-assisted laser desorption/ionization mass spectrometry can distinguish breast cancer patients from cancer-free controls.

Herein, we report that breast cancer (BC) patients can be distinguished from cancer-free (NC) controls by serum immunoglobulin G (IgG) crystallizable fragment (Fc) region N-glycosylation profiling using matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS). Recently, there has been much progress in the field of tumor immunology. However, to date, the role and biomarker potential of IgG Fc region N-glycosylation, which affects the function of antibodies, have not been examined in BC. In the present study, we profiled serum IgG Fc region N-glycans in BC patients (N = 90) and NC controls (N = 54) using MALDI-MS. An IgG Fc region N-glycan-based multiple logistic regression model was produced which could distinguish BC patients from NC controls (area under the receiver operative characteristic curve = 0.874). Furthermore, stage 0 patients could also be distinguished using this model. These results suggest that an unknown humoral factor or soluble mediator affects IgGs from the earliest stage of breast cancer, and also suggests that IgG Fc region N-glycosylation may play a role in tumor biology. Although further investigation is required, our findings are the evidence that IgG N-glycan profiling has the potential to be used as a breast cancer biomarker and may provide the insights into tumor immunology.

GlycanAnalysis Plug-in: a database search tool for N-glycan structures using mass spectrometry.

UNLABELLED: Tandem mass spectrometry (MS/MS or MS(n)) is a potent technique for characterizing N-glycan structures. GlycanAnalysis searches a glycan database to support the identification of glycan structures from MS/MS spectra. It also calculates diagnostic ions of glycan structures registered in a glycan database (GlycomeDB or KEGG GLYCAN) and searches for MS/MS spectra of N-glycans that match diagnostic ions to determine the structures. This program functions as a plug-in for Mass++, a freeware mass spectrum visualization and analysis program. AVAILABILITY AND IMPLEMENTATION: The executable files of Mass++ are available for free at http://www.first-ms3d.jp/english/. The GlycanAnalysis plug-in is included in the standard package of Mass++ for Windows. CONTACT: k-morimt@shimadzu.co.jp or nishikaz@shimadzu.co.jp or acyshzw@shimadzu.co.jp SUPPLEMENTARY INFORMATION: Supplementary material are available at Bioinformatics online.

Computational survey of sequence specificity for protein terminal tags covering nine organisms and its application to protein identification.

In 1998, Wilkins et al. (J. Mol. Biol. 1998, 278, 599-608) reported high specificity in terminal regions (terminal tags) of 15 519 proteins from five organisms and proposed a methodology for identifying proteins by terminal tags. However, their examined sequence data were not based on complete genome sequences. Here, we examined current proteome data (217 249 entries from UniProt 2013_6 complete/reference proteome for nine organisms including human) in terms of the specificity of terminal tags and their computational annotation. One example from the results indicated that the specificity of N-terminal tags plateaued at 28% at a length of six residues for human; even when using both N- and C-terminal tags, specificity was merely 66%. In order to determine the cause of these low specificities, the annotation of proteins sharing terminal tags with other proteins was examined. The results suggested that a large majority were phylogenetically or functionally related, whereas nonrelated proteins sharing terminal tags made up less than 1% of human proteome data. On the basis of these findings, we constructed the terminal tag sequence database ProteinCarta (http://ms3d.jp/software/proteincarta/), which includes all terminal tags of proteomes from the nine organisms analyzed here, in order to confirm the specificity of terminal tags and to identify the parent protein.

A simple peak detection and label-free quantitation algorithm for chromatography-mass spectrometry.

BACKGROUND: Label-free quantitation of mass spectrometric data is one of the simplest and least expensive methods for differential expression profiling of proteins and metabolites. The need for high accuracy and performance computational label-free quantitation methods is still high in the biomarker and drug discovery research field. However, recent most advanced types of LC-MS generate huge amounts of analytical data with high scan speed, high accuracy and resolution, which is often impossible to interpret manually. Moreover, there are still issues to be improved for recent label-free methods, such as how to reduce false positive/negatives of the candidate peaks, how to expand scalability and how to enhance and automate data processing. AB3D (A simple label-free quantitation algorithm for Biomarker Discovery in Diagnostics and Drug discovery using LC-MS) has addressed these issues and has the capability to perform label-free quantitation using MS1 for proteomics study. RESULTS: We developed an algorithm called AB3D, a label free peak detection and quantitative algorithm using MS1 spectral data. To test our algorithm, practical applications of AB3D for LC-MS data sets were evaluated using 3 datasets. Comparisons were then carried out between widely used software tools such as MZmine 2, MSight, SuperHirn, OpenMS and our algorithm AB3D, using the same LC-MS datasets. All quantitative results were confirmed manually, and we found that AB3D could properly identify and quantify known peptides with fewer false positives and false negatives compared to four other existing software tools using either the standard peptide mixture or the real complex biological samples of Bartonella quintana (strain JK31). Moreover, AB3D showed the best reliability by comparing the variability between two technical replicates using a complex peptide mixture of HeLa and BSA samples. For performance, the AB3D algorithm is about 1.2 - 15 times faster than the four other existing software tools. CONCLUSIONS: AB3D is a simple and fast algorithm for label-free quantitation using MS1 mass spectrometry data for large scale LC-MS data analysis with higher true positive and reasonable false positive rates. Furthermore, AB3D demonstrated the best reproducibility and is about 1.2- 15 times faster than those of existing 4 software tools.

The C-terminal fragment of prostate-specific antigen, a 2331 Da peptide, as a new urinary pathognomonic biomarker candidate for diagnosing prostate cancer.

BACKGROUND AND OBJECTIVES: Prostate cancer (PCa) is one of the most common cancers and leading cause of cancer-related deaths in men. Mass screening has been carried out since the 1990s using prostate-specific antigen (PSA) levels in the serum as a PCa biomarker. However, although PSA is an excellent organ-specific marker, it is not a cancer-specific marker. Therefore, the aim of this study was to discover new biomarkers for the diagnosis of PCa. MATERIALS AND METHODS: We focused on urine samples voided following prostate massage (digital rectal examination [DRE]) and conducted a peptidomic analysis of these samples using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF/MS(n)). Urinary biomaterials were concentrated and desalted using CM-Sepharose prior to the following analyses being performed by MALDI-TOF/MS(n): 1) differential analyses of mass spectra; 2) determination of amino acid sequences; and 3) quantitative analyses using a stable isotope-labeled internal standard. RESULTS: Multivariate analysis of the MALDI-TOF/MS mass spectra of urinary extracts revealed a 2331 Da peptide in urine samples following DRE. This peptide was identified as a C-terminal PSA fragment composed of 19 amino acid residues. Moreover, quantitative analysis of the relationship between isotope-labeled synthetic and intact peptides using MALDI-TOF/MS revealed that this peptide may be a new pathognomonic biomarker candidate that can differentiate PCa patients from non-cancer subjects. CONCLUSION: The results of the present study indicate that the 2331 Da peptide fragment of PSA may become a new pathognomonic biomarker for the diagnosis of PCa. A further large-scale investigation is currently underway to assess the possibility of using this peptide in the early detection of PCa.

Mass++: A Visualization and Analysis Tool for Mass Spectrometry.

We have developed Mass++, a plug-in style visualization and analysis tool for mass spectrometry. Its plug-in style enables users to customize it and to develop original functions. Mass++ has several kinds of plug-ins, including rich viewers and analysis methods for proteomics and metabolomics. Plug-ins for supporting vendors' raw data are currently available; hence, Mass++ can read several data formats. Mass++ is both a desktop tool and a software development platform. Original functions can be developed without editing the Mass++ source code. Here, we present this tool's capability to rapidly analyze MS data and develop functions by providing examples of label-free quantitation and implementing plug-ins or scripts. Mass++ is freely available at http://www.first-ms3d.jp/english/ .

Peptide Peak Detection for Low Resolution MALDI-TOF Mass Spectrometry.

A new peak detection method has been developed for rapid selection of peptide and its fragment ion peaks for protein identification using tandem mass spectrometry. The algorithm applies classification of peak intensities present in the defined mass range to determine the noise level. A threshold is then given to select ion peaks according to the determined noise level in each mass range. This algorithm was initially designed for the peak detection of low resolution peptide mass spectra, such as matrix-assisted laser desorption/ionization Time-of-Flight (MALDI-TOF) mass spectra. But it can also be applied to other type of mass spectra. This method has demonstrated obtaining a good rate of number of real ions to noises for even poorly fragmented peptide spectra. The effect of using peak lists generated from this method produces improved protein scores in database search results. The reliability of the protein identifications is increased by finding more peptide identifications. This software tool is freely available at the Mass++ home page (http://www.first-ms3d.jp/english/achievement/software/).

MALDI mass spectrometry-based sequence analysis of arginine-containing glycopeptides: improved fragmentation of glycan and peptide chains by modifying arginine residue.

This paper describes an improved method for the sequence analysis of Arg-containing glycopeptide by MALDI mass spectrometry (MS). The method uses amino group derivatization (4-aza-6-(2,6-dimethyl-1-piperidinyl)-5-oxohexanoic acid N-succinimidyl ester) and removal (carboxypeptidase B) or modification (peptidylarginine deiminase 4) of the arginine residue of the peptide. The derivatization attaches a basic tertiary amine moiety onto the peptides, and the enzymatic treatment removes or modifies the arginine residue. Fragmentation of the resulting glycopeptide under low-energy collision-induced dissociation yielded a simplified ion series of both the glycan and the peptide that can facilitate their sequencing. The feasibility of the method was studied using α1 -acid glycoprotein-derived N-linked glycopeptides, and glycan and peptide in each glycopeptide were successfully sequenced by MALDI tandem MS (MS/MS).

Semi-quantitative analyses of metabolic systems of human colon cancer metastatic xenografts in livers of superimmunodeficient NOG mice.

Analyses of energy metabolism in human cancer have been difficult because of rapid turnover of the metabolites and difficulties in reducing time for collecting clinical samples under surgical procedures. Utilization of xenograft transplantation of human-derived colon cancer HCT116 cells in spleens of superimmunodeficient NOD/SCID/IL-2Rγ(null) (NOG) mice led us to establish an experimental model of hepatic micrometastasis of the solid tumor, whereby analyses of the tissue sections collected by snap-frozen procedures through newly developed microscopic imaging mass spectrometry (MIMS) revealed distinct spatial distribution of a variety of metabolites. To perform intergroup comparison of the signal intensities of metabolites among different tissue sections collected from mice in fed states, we combined matrix-assisted laser desorption/ionization time-of-flight imaging mass spectrometry (MALDI-TOF-IMS) and capillary electrophoresis-mass spectrometry (CE-MS), to determine the apparent contents of individual metabolites in serial tissue sections. The results indicated significant elevation of ATP and energy charge in both metastases and the parenchyma of the tumor-bearing livers. To note were significant increases in UDP-N-acetyl hexosamines, and reduced and oxidized forms of glutathione in the metastatic foci versus the liver parenchyma. These findings thus provided a potentially important method for characterizing the properties of metabolic systems of human-derived cancer and the host tissues in vivo.

Integrated mass spectrometry-based analysis of plasma glycoproteins and their glycan modifications.

We present a protocol for the identification of glycosylated proteins in plasma followed by elucidation of their individual glycan compositions. The study of glycoproteins by mass spectrometry is usually based on cleavage of glycans followed by separate analysis of glycans and deglycosylated proteins, which limits the ability to derive glycan compositions for individual glycoproteins. The methodology described here consists of 2D HPLC fractionation of intact proteins and liquid chromatography-multistage tandem mass spectrometry (LC-MS/MS(n)) analysis of digested protein fractions. Protein samples are separated by 1D anion-exchange chromatography (AEX) with an eight-step salt elution. Protein fractions from each of the eight AEX elution steps are transferred onto the 2D reversed-phase column to further separate proteins. A digital ion trap mass spectrometer with a wide mass range is then used for LC-MS/MS(n) analysis of intact glycopeptides from the 2D HPLC fractions. Both peptide and oligosaccharide compositions are revealed by analysis of the ion fragmentation patterns of glycopeptides with an intact glycopeptide analysis pipeline.

Development of imaging mass spectrometry (IMS) dataset extractor software, IMS convolution.

Imaging mass spectrometry (IMS) is a powerful tool for detecting and visualizing biomolecules in tissue sections. The technology has been applied to several fields, and many researchers have started to apply it to pathological samples. However, it is very difficult for inexperienced users to extract meaningful signals from enormous IMS datasets, and the procedure is time-consuming. We have developed software, called IMS Convolution with regions of interest (ROI), to automatically extract meaningful signals from IMS datasets. The processing is based on the detection of common peaks within the ordered area in the IMS dataset. In this study, the IMS dataset from a mouse eyeball section was acquired by a mass microscope that we recently developed, and the peaks extracted by manual and automatic procedures were compared. The manual procedure extracted 16 peaks with higher intensity in mass spectra averaged in whole measurement points. On the other hand, the automatic procedure using IMS Convolution easily and equally extracted peaks without any effort. Moreover, the use of ROIs with IMS Convolution enabled us to extract the peak on each ROI area, and all of the 16 ion images on mouse eyeball tissue were from phosphatidylcholine species. Therefore, we believe that IMS Convolution with ROIs could automatically extract the meaningful peaks from large-volume IMS datasets for inexperienced users as well as for researchers who have performed the analysis.

Salamander retina phospholipids and their localization by MALDI imaging mass spectrometry at cellular size resolution.

Salamander large cells facilitated identification and localization of lipids by MALDI imaging mass spectrometry. Salamander retina lipid extract showed similarity with rodent retina lipid extract in phospholipid content and composition. Like rodent retina section, distinct layer distributions of phospholipids were observed in the salamander retina section. Phosphatidylcholines (PCs) composing saturated and monounsaturated fatty acids (PC 32:0, PC 32:1, and PC 34:1) were detected mainly in the outer and inner plexiform layers (OPL and IPL), whereas PCs containing polyunsaturated fatty acids (PC 36:4, PC 38:6, and PC 40:6) composed the inner segment (IS) and outer segment (OS). The presence of PCs containing polyunsaturated fatty acids in the OS layer implied that these phospholipids form flexible lipid bilayers, which facilitate phototransduction process occurring in the rhodopsin rich OS layer. Distinct distributions and relative signal intensities of phospholipids also indicated their relative abundance in a particular cell or a cell part. Using salamander large cells, a single cell level localization and identification of biomolecules could be achieved by MALDI imaging mass spectrometry.

Visualization of the cell-selective distribution of PUFA-containing phosphatidylcholines in mouse brain by imaging mass spectrometry.

Previous studies have shown that MALDI-imaging mass spectrometry (IMS) can be used to visualize the distribution of various biomolecules, especially lipids, in the cells and tissues. In this study, we report the cell-selective distribution of PUFA-containing glycerophospholipids (GPLs) in the mouse brain. We established a practical experimental procedure for the IMS of GPLs. We demonstrated that optimization of the composition of the matrix solution and spectrum normalization to the total ion current (TIC) is critical. Using our procedure, we simultaneously differentiated and visualized the localizations of specific molecular species of GPLs in mouse brain sections. The results showed that PUFA-containing phosphatidylcholines (PCs) were distributed in a cell-selective manner: arachidonic acid- and docosahexaenoic acid-containing PCs were seen in the hippocampal neurons and cerebellar Purkinje cells, respectively. Furthermore, these characteristic localizations of PUFA-PCs were formed during neuronal maturation. The phenomenon of brain cell-selective production of specific PUFA-GPLs will help elucidate the potential physiological functions of PUFAs in specific brain regions.

Hierarchical Bayesian estimation for MEG inverse problem.

Source current estimation from MEG measurement is an ill-posed problem that requires prior assumptions about brain activity and an efficient estimation algorithm. In this article, we propose a new hierarchical Bayesian method introducing a hierarchical prior that can effectively incorporate both structural and functional MRI data. In our method, the variance of the source current at each source location is considered an unknown parameter and estimated from the observed MEG data and prior information by using the Variational Bayesian method. The fMRI information can be imposed as prior information on the variance distribution rather than the variance itself so that it gives a soft constraint on the variance. A spatial smoothness constraint, that the neural activity within a few millimeter radius tends to be similar due to the neural connections, can also be implemented as a hierarchical prior. The proposed method provides a unified theory to deal with the following three situations: (1) MEG with no other data, (2) MEG with structural MRI data on cortical surfaces, and (3) MEG with both structural MRI and fMRI data. We investigated the performance of our method and conventional linear inverse methods under these three conditions. Simulation results indicate that our method has better accuracy and spatial resolution than the conventional linear inverse methods under all three conditions. It is also shown that accuracy of our method improves as MRI and fMRI information becomes available. Simulation results demonstrate that our method appropriately resolves the inverse problem even if fMRI data convey inaccurate information, while the Wiener filter method is seriously deteriorated by inaccurate fMRI information.