Method Development for Unknown Compound Identification in Mass Spectrometry-based Untargeted Metabolomics

Metabolomics is the study of the collective small molecules within a biological system (cell, tissue or whole organism). The study of these small molecules (metabolites) gives a close measure of the phenotype, giving insight into an organism’s physiological and biochemical state at the time of sample collection. Thus, metabolomics has emerged as a highly attractive field employed to study normal physiology and changes in physiology due to natural diversity, genetic mutations, and disease state(s). Liquid chromatography coupled to mass spectrometry (LC-MS) has become the main analytical platform in the field of metabolomics due to its unmatched combination of molecular coverage, sensitivity, and specificity. A typical mass spectrometry-based untargeted metabolomics workflow involves: experimental design, sample collection, metabolite extraction, mass spectrometry data collection, and analysis. Upstream steps such as metabolite extraction determine the complexity of data interpretation, and while improvements in bioinformatic approaches and available reference databases have helped to address this issue, only a small percentage (<2%) of detected mass spectral features can be confidently identified. To address this major bottleneck of compound identification, I have utilized a highly efficient Taguchi design of experiments (DoE) approach to optimize a sequential Caenorhabditis elegans(C. elegans) nonpolar and polar extraction using reverse phase (RP) and hydrophilic interaction liquid chromatography (HILIC) coupled to high resolution mass spectrometry (HRMS), respectively. This resulted in a significant decrease in the number of samples required to optimize the extraction, while maintaining a data structure that allowed for interactions between study factors to be investigated. Using this optimized extraction, semi-preparative HPLC was used to create a metabolome fraction library of C. elegans laboratory reference strain (PD1074). Aliquots of each fraction was analyzed by LC-MS/MS, NMR, and additional aliquots were stored for ion mobility and ultra-high resolution FT-ICR-MS. Statistical methods such as Statistical Total Correlation Spectroscopy (STOCSY), molecular networking (GNPS), and retention time alignment (MetabCombiner) were used to bridge these highly complementary datasets and improve our compound identification strategy. While metabolite ID remaines the key bottleneck in metabolomics, the work showcased herein demonstrates the importance of each step of the metabolomics workflow, necessity to optimize these steps, and the strengths complementary analytical techniques bring to the compound ID table.