Development and in silico evaluation of large-scale metabolite identification methods using functional group detection for metabolomics

md_harmonize: A Python Package for Atom-Level Harmonization of Public Metabolic Databases

A major challenge to integrating public metabolic resources is the use of different nomenclatures by individual databases. This paper presents md_harmonize, an open-source Python package for harmonizing compounds and metabolic reactions across various metabolic databases. The md_harmonize package utilizes a neighborhood-specific graph coloring method for generating a unique identifier for each compound via atom identifiers based on a compound's chemical structure. The resulting harmonized compounds and reactions can be used for various downstream analyses, including the construction of atom-resolved metabolic networks and models for metabolic flux analysis. Parts of the md_harmonize package have been optimized using a variety of computational techniques to allow certain NP-complete problems handled by the software to be tractable for these specific use-cases. The software is available on GitHub and through the Python Package Index, with end-user documentation hosted on GitHub Pages.

Benchmark Dataset for Training Machine Learning Models to Predict the Pathway Involvement of Metabolites

Metabolic pathways are a human-defined grouping of life sustaining biochemical reactions, metabolites being both the reactants and products of these reactions. But many public datasets include identified metabolites whose pathway involvement is unknown, hindering metabolic interpretation. To address these shortcomings, various machine learning models, including those trained on data from the Kyoto Encyclopedia of Genes and Genomes (KEGG), have been developed to predict the pathway involvement of metabolites based on their chemical descriptions; however, these prior models are based on old metabolite KEGG-based datasets, including one benchmark dataset that is invalid due to the presence of over 1500 duplicate entries. Therefore, we have developed a new benchmark dataset derived from the KEGG following optimal standards of scientific computational reproducibility and including all source code needed to update the benchmark dataset as KEGG changes. We have used this new benchmark dataset with our atom coloring methodology to develop and compare the performance of Random Forest, XGBoost, and multilayer perceptron with autoencoder models generated from our new benchmark dataset. Best overall weighted average performance across 1000 unique folds was an F1 score of 0.8180 and a Matthews correlation coefficient of 0.7933, which was provided by XGBoost binary classification models for 11 KEGG-defined pathway categories.

Benchmark dataset for training machine learning models to predict the pathway involvement of metabolites

Metabolic pathways are a human-defined grouping of life sustaining biochemical reactions, metabolites being both the reactants and products of these reactions. But many public datasets include identified metabolites whose pathway involvement is unknown, hindering metabolic interpretation. To address these shortcomings, various machine learning models, including those trained on data from the Kyoto Encyclopedia of Genes and Genomes (KEGG), have been developed to predict the pathway involvement of metabolites based on their chemical descriptions; however, these prior models are based on old metabolite KEGG-based datasets, including one benchmark dataset that is invalid due to the presence of over 1500 duplicate entries. Therefore, we have developed a new benchmark dataset derived from the KEGG following optimal standards of scientific computational reproducibility and including all source code needed to update the benchmark dataset as KEGG changes. We have used this new benchmark dataset with our atom coloring methodology to develop and compare the performance of Random Forest, XGBoost, and multilayer perceptron with autoencoder models generated from our new benchmark dataset. Best overall weighted average performance across 1000 unique folds was an F1-score of 0.8180 and Matthews correlation coefficient of 0.7933, which was provided by XGBoost binary classification models for 11 KEGG-defined pathway categories.

Untargeted Lipidomics of Non-Small Cell Lung Carcinoma Demonstrates Differentially Abundant Lipid Classes in Cancer vs. Non-Cancer Tissue

Lung cancer remains the leading cause of cancer death worldwide and non-small cell lung carcinoma (NSCLC) represents 85% of newly diagnosed lung cancers. In this study, we utilized our untargeted assignment tool Small Molecule Isotope Resolved Formula Enumerator (SMIRFE) and ultra-high-resolution Fourier transform mass spectrometry to examine lipid profile differences between paired cancerous and non-cancerous lung tissue samples from 86 patients with suspected stage I or IIA primary NSCLC. Correlation and co-occurrence analysis revealed significant lipid profile differences between cancer and non-cancer samples. Further analysis of machine-learned lipid categories for the differentially abundant molecular formulas identified a high abundance sterol, high abundance and high m/z sphingolipid, and low abundance glycerophospholipid metabolic phenotype across the NSCLC samples. At the class level, higher abundances of sterol esters and lower abundances of cardiolipins were observed suggesting altered stearoyl-CoA desaturase 1 (SCD1) or acetyl-CoA acetyltransferase (ACAT1) activity and altered human cardiolipin synthase 1 or lysocardiolipin acyltransferase activity respectively, the latter of which is known to confer apoptotic resistance. The presence of a shared metabolic phenotype across a variety of genetically distinct NSCLC subtypes suggests that this phenotype is necessary for NSCLC development and may result from multiple distinct genetic lesions. Thus, targeting the shared affected pathways may be beneficial for a variety of genetically distinct NSCLC subtypes.

Hierarchical Harmonization of Atom-Resolved Metabolic Reactions across Metabolic Databases

Metabolic models have been proven to be useful tools in system biology and have been successfully applied to various research fields in a wide range of organisms. A relatively complete metabolic network is a prerequisite for deriving reliable metabolic models. The first step in constructing metabolic network is to harmonize compounds and reactions across different metabolic databases. However, effectively integrating data from various sources still remains a big challenge. Incomplete and inconsistent atomistic details in compound representations across databases is a very important limiting factor. Here, we optimized a subgraph isomorphism detection algorithm to validate generic compound pairs. Moreover, we defined a set of harmonization relationship types between compounds to deal with inconsistent chemical details while successfully capturing atom-level characteristics, enabling a more complete enabling compound harmonization across metabolic databases. In total, 15,704 compound pairs across KEGG (Kyoto Encyclopedia of Genes and Genomes) and MetaCyc databases were detected. Furthermore, utilizing the classification of compound pairs and EC (Enzyme Commission) numbers of reactions, we established hierarchical relationships between metabolic reactions, enabling the harmonization of 3856 reaction pairs. In addition, we created and used atom-specific identifiers to evaluate the consistency of atom mappings within and between harmonized reactions, detecting some consistency issues between the reaction and compound descriptions in these metabolic databases.

Atom Identifiers Generated by a Neighborhood-Specific Graph Coloring Method Enable Compound Harmonization across Metabolic Databases

Metabolic flux analysis requires both a reliable metabolic model and reliable metabolic profiles in characterizing metabolic reprogramming. Advances in analytic methodologies enable production of high-quality metabolomics datasets capturing isotopic flux. However, useful metabolic models can be difficult to derive due to the lack of relatively complete atom-resolved metabolic networks for a variety of organisms, including human. Here, we developed a neighborhood-specific graph coloring method that creates unique identifiers for each atom in a compound facilitating construction of an atom-resolved metabolic network. What is more, this method is guaranteed to generate the same identifier for symmetric atoms, enabling automatic identification of possible additional mappings caused by molecular symmetry. Furthermore, a compound coloring identifier derived from the corresponding atom coloring identifiers can be used for compound harmonization across various metabolic network databases, which is an essential first step in network integration. With the compound coloring identifiers, 8865 correspondences between KEGG (Kyoto Encyclopedia of Genes and Genomes) and MetaCyc compounds are detected, with 5451 of them confirmed by other identifiers provided by the two databases. In addition, we found that the Enzyme Commission numbers (EC) of reactions can be used to validate possible correspondence pairs, with 1848 unconfirmed pairs validated by commonality in reaction ECs. Moreover, we were able to detect various issues and errors with compound representation in KEGG and MetaCyc databases by compound coloring identifiers, demonstrating the usefulness of this methodology for database curation.

Deriving Lipid Classification Based on Molecular Formulas

Despite instrument and algorithmic improvements, the untargeted and accurate assignment of metabolites remains an unsolved problem in metabolomics. New assignment methods such as our SMIRFE algorithm can assign elemental molecular formulas to observed spectral features in a highly untargeted manner without orthogonal information from tandem MS or chromatography. However, for many lipidomics applications, it is necessary to know at least the lipid category or class that is associated with a detected spectral feature to derive a biochemical interpretation. Our goal is to develop a method for robustly classifying elemental molecular formula assignments into lipid categories for an application to SMIRFE-generated assignments. Using a Random Forest machine learning approach, we developed a method that can predict lipid category and class from SMIRFE non-adducted molecular formula assignments. Our methods achieve high average predictive accuracy (>90%) and precision (>83%) across all eight of the lipid categories in the LIPIDMAPS database. Classification performance was evaluated using sets of theoretical, data-derived, and artifactual molecular formulas. Our methods enable the lipid classification of non-adducted molecular formula assignments generated by SMIRFE without orthogonal information, facilitating the biochemical interpretation of untargeted lipidomics experiments. This lipid classification appears insufficient for validating single-spectrum assignments, but could be useful in cross-spectrum assignment validation.

Small Molecule Isotope Resolved Formula Enumeration: A Methodology for Assigning Isotopologues and Metabolite Formulas in Fourier Transform Mass Spectra

Improvements in Fourier transform mass spectrometry (FT-MS) enable increasingly more complex experiments in the field of metabolomics. What is directly detected in FT-MS spectra are spectral features (peaks) that correspond to sets of adducted and charged forms of specific molecules in the sample. The robust assignment of these features is an essential step for MS-based metabolomics experiments, but the sheer complexity of what is detected and a variety of analytically introduced variance, errors, and artifacts has hindered the systematic analysis of complex patterns of observed peaks with respect to isotope content. We have developed a method called SMIRFE that detects small biomolecules and determines their elemental molecular formula (EMF) using detected sets of isotopologue peaks sharing the same EMF. SMIRFE does not use a database of known metabolite formulas; instead a nearly comprehensive search space of all isotopologues within a mass range is constructed and used for assignment. This search space can be tailored for different isotope labeling patterns expected in different stable isotope tracing experiments. Using consumer-level computing equipment, a large search space of 2000 Da was constructed, and assignment performance was evaluated and validated using verified assignments on a pair of peak lists derived from spectra containing unlabeled and ¹⁵N-labeled versions of amino acids derivatized using ethylchloroformate. SMIRFE identified 18 of 18 predicted derivatized EMFs, and each assignment was evaluated statistically and assigned an e-value representing the probability to occur by chance.

Design and Success of a 21st Century Cancer Education Program at the University of Louisville

Cancer incidence, morbidity, and mortality in the Commonwealth of Kentucky are among the highest in the nation. The University of Louisville was the recipient of a National Cancer Institute (NCI)-funded cancer education program grant in 1975 under the leadership of Dr. Norbert Burzynski. A new and totally redesigned performance-based University of Louisville Cancer Education Program was funded by NCI in 2011 to recruit and motivate outstanding undergraduate and health professional students to pursue further training and careers in cancer research. Here, we describe the strategy, design, methods, implementation, and accomplishments of our twenty-first century performance-based cancer education program. Our program will meet or exceed all of its 5-year performance goals, including the total number students (n = 156) and under-represented minorities (n = 53) who successfully completed the program under the mentorship of cancer research-intensive faculty members of the James Graham Brown Cancer Center (JGBCC). The mentored research program is complemented with professional development and enhancement activities, including cancer research seminars presented by faculty members actively engaged in research centered on the diagnosis, treatment or prevention of cancer, creation of individual career development plans, exploration of cancer research careers, and acquisition of professionalism skills. Student interests towards cancer research significantly increased after completion of the program compared to baseline (P = 0.02). Based on quantitative and qualitative analysis of various components of the curricula, the trainees favor practical, engaging, and interactive activities aligned within professional career goals and objectives. For instance, the trainees prefer two 30-min small group discussions on "Navigating Careers in Cancer Research" with faculty, professional students, and program alumni. Future updates to the program include new activities that capitalize on the cross-disciplinary background of our mentors and trainees as well as a team-based approach to professional development. Our cancer education program will continue to enhance the professional development of the next generation of cancer scientists and clinicians.

Perspectives and expectations in structural bioinformatics of metalloproteins

Recent papers highlight the presence of large numbers of compressed angles in metal ion coordination geometries for metalloprotein entries in the worldwide Protein Data Bank, due mainly to multidentate coordination. The prevalence of these compressed angles has raised the controversial idea that significantly populated aberrant or even novel coordination geometries may exist. Some of these papers have undergone severe criticism, apparently due to views held that only canonical coordination geometries exist in significant numbers. While criticism of controversial ideas is warranted and to be expected, we believe that a line was crossed where unfair criticism was put forth to discredit an inconvenient result that compressed angles exist in large numbers, which does not support the dogmatic canonical coordination geometry view. We present a review of the major controversial results and their criticisms, pointing out both good suggestions that have been incorporated in new analyses, but also unfair criticism that was put forth to support a particular view. We also suggest that better science is enabled through: (i) a more collegial and collaborative approach in future critical reviews and (ii) the requirement for a description of methods and data including source code and visualizations that enables full reproducibility of results. Proteins 2017; 85:938-944. © 2016 Wiley Periodicals, Inc.

Preclinical models for interrogating drug action in human cancers using Stable Isotope Resolved Metabolomics (SIRM)

AIMS

In this review we compare the advantages and disadvantages of different model biological systems for determining the metabolic functions of cells in complex environments, how they may change in different disease states, and respond to therapeutic interventions.

BACKGROUND

All preclinical drug-testing models have advantages and drawbacks. We compare and contrast established cell, organoid and animal models with ex vivo organ or tissue culture and in vivo human experiments in the context of metabolic readout of drug efficacy. As metabolism reports directly on the biochemical state of cells and tissues, it can be very sensitive to drugs and/or other environmental changes. This is especially so when metabolic activities are probed by stable isotope tracing methods, which can also provide detailed mechanistic information on drug action. We have developed and been applying Stable Isotope-Resolved Metabolomics (SIRM) to examine metabolic reprogramming of human lung cancer cells in monoculture, in mouse xenograft/explant models, and in lung cancer patients in situ (Lane et al. 2011; T. W. Fan et al. 2011; T. W-M. Fan et al. 2012; T. W. Fan et al. 2012; Xie et al. 2014b; Ren et al. 2014a; Sellers et al. 2015b). We are able to determine the influence of the tumor microenvironment using these models. We have now extended the range of models to fresh human tissue slices, similar to those originally described by O. Warburg (Warburg 1923), which retain the native tissue architecture and heterogeneity with a paired benign versus cancer design under defined cell culture conditions. This platform offers an unprecedented human tissue model for preclinical studies on metabolic reprogramming of human cancer cells in their tissue context, and response to drug treatment (Xie et al. 2014a). As the microenvironment of the target human tissue is retained and individual patient's response to drugs is obtained, this platform promises to transcend current limitations of drug selection for clinical trials or treatments.

CONCLUSIONS AND FUTURE WORK

Development of ex vivo human tissue and animal models with humanized organs including bone marrow and liver show considerable promise for analyzing drug responses that are more relevant to humans. Similarly using stable isotope tracer methods with these improved models in advanced stages of the drug development pipeline, in conjunction with tissue biopsy is expected significantly to reduce the high failure rate of experimental drugs in Phase II and III clinical trials.

Metabolome-scale de novo pathway reconstruction using regioisomer-sensitive graph alignments

MOTIVATION

Recent advances in mass spectrometry and related metabolomics technologies have enabled the rapid and comprehensive analysis of numerous metabolites. However, biosynthetic and biodegradation pathways are only known for a small portion of metabolites, with most metabolic pathways remaining uncharacterized.

RESULTS

In this study, we developed a novel method for supervised de novo metabolic pathway reconstruction with an improved graph alignment-based approach in the reaction-filling framework. We proposed a novel chemical graph alignment algorithm, which we called PACHA (Pairwise Chemical Aligner), to detect the regioisomer-sensitive connectivities between the aligned substructures of two compounds. Unlike other existing graph alignment methods, PACHA can efficiently detect only one common subgraph between two compounds. Our results show that the proposed method outperforms previous descriptor-based methods or existing graph alignment-based methods in the enzymatic reaction-likeness prediction for isomer-enriched reactions. It is also useful for reaction annotation that assigns potential reaction characteristics such as EC (Enzyme Commission) numbers and PIERO (Enzymatic Reaction Ontology for Partial Information) terms to substrate-product pairs. Finally, we conducted a comprehensive enzymatic reaction-likeness prediction for all possible uncharacterized compound pairs, suggesting potential metabolic pathways for newly predicted substrate-product pairs.

Using fragmentation trees and mass spectral trees for identifying unknown compounds in metabolomics

Identification of unknown metabolites is the bottleneck in advancing metabolomics, leaving interpretation of metabolomics results ambiguous. The chemical diversity of metabolism is vast, making structure identification arduous and time consuming. Currently, comprehensive analysis of mass spectra in metabolomics is limited to library matching, but tandem mass spectral libraries are small compared to the large number of compounds found in the biosphere, including xenobiotics. Resolving this bottleneck requires richer data acquisition and better computational tools. Multi-stage mass spectrometry (MSn) trees show promise to aid in this regard. Fragmentation trees explore the fragmentation process, generate fragmentation rules and aid in sub-structure identification, while mass spectral trees delineate the dependencies in multi-stage MS of collision-induced dissociations. This review covers advancements over the past 10 years as a tool for metabolite identification, including algorithms, software and databases used to build and to implement fragmentation trees and mass spectral annotations.

Nutrition and metabolic correlates of obesity and inflammation: clinical considerations

Since 1980, the global prevalence of obesity has doubled; in the United States, it has almost tripled. Billions of people are overweight and obese; the WHO reports that >65% of the world's population die of diseases related to overweight rather than underweight. Obesity is a complex disease that can be studied from "metropolis to metabolite"—that is, beginning at the policy and the population level through epidemiology and intervention studies; to bench work including preclinical models, tissue, and cell culture studies; to biochemical assays; and to metabolomics. Metabolomics is the next research frontier because it provides a real-time snapshot of biochemical building blocks and products of cellular processes. This report comments on practical considerations when conducting metabolomics research. The pros and cons and important study design concerns are addressed to aid in increasing metabolomics research in the United States. The link between metabolism and inflammation is an understudied phenomenon that has great potential to transform our understanding of immunometabolism in obesity, diabetes, cancer, and other diseases; metabolomics promises to be an important tool in understanding the complex relations between factors contributing to such diseases.

Systems-based approaches to unravel multi-species microbial community functioning

Some of the most transformative discoveries promising to enable the resolution of this century's grand societal challenges will most likely arise from environmental science and particularly environmental microbiology and biotechnology. Understanding how microbes interact in situ, and how microbial communities respond to environmental changes remains an enormous challenge for science. Systems biology offers a powerful experimental strategy to tackle the exciting task of deciphering microbial interactions. In this framework, entire microbial communities are considered as metaorganisms and each level of biological information (DNA, RNA, proteins and metabolites) is investigated along with in situ environmental characteristics. In this way, systems biology can help unravel the interactions between the different parts of an ecosystem ultimately responsible for its emergent properties. Indeed each level of biological information provides a different level of characterisation of the microbial communities. Metagenomics, metatranscriptomics, metaproteomics, metabolomics and SIP-omics can be employed to investigate collectively microbial community structure, potential, function, activity and interactions. Omics approaches are enabled by high-throughput 21st century technologies and this review will discuss how their implementation has revolutionised our understanding of microbial communities.