The predicted liquid chromatographic retention times (RTs) of small molecules are not accurate enough for wide adoption in structural identification. In this study, we used the graph neural network to predict the retention time (GNN-RT) from structures of small molecules directly without the requirement of molecular descriptors. The predicted accuracy of GNN-RT was compared with random forests (RFs), Bayesian ridge regression, convolutional neural network (CNN), and a deep-learning regression model (DLM) on a METLIN small molecule retention time (SMRT) dataset. GNN-RT achieved the highest predicting accuracy with a mean relative error of 4.9% and a median relative error of 3.2%. Furthermore, the SMRT-trained GNN-RT model can be transferred to the same type of chromatographic systems easily. The predicted RT is valuable for structural identification in complementary to tandem mass spectra and can be used to assist in the identification of compounds. The results indicate that GNN-RT is a promising method to predict the RT for liquid chromatography and improve the accuracy of structural identification for small molecules.

Electron ionization−mass spectrometry (EI-MS) hyphenated to gas chromatography (GC) is the workhorse for analyzing volatile compounds in complex samples. The spectral matching method can only identify compounds within the spectral database. In response, we present a deep-learning-based approach (DeepEI) for structure elucidation of an unknown compound with its EI-MS spectrum. DeepEI employs deep neural networks to predict molecular fingerprints from an EI-MS spectrum and searches the molecular structure database with the predicted fingerprints. We evaluated DeepEI with MassBank spectra, and the results indicate DeepEI is an effective identification method. In addition, DeepEI can work cooperatively with database spectral matching and NEIMS (fingerprint to spectrum method) to improve identification accuracy.

No abstract

Tandem mass spectrometry (MS/MS) is the workhorse for structural annotation of metabolites, because it can provide abundance of structural information. Currently, metabolite identification mainly relies on querying experimental spectra against public or in-house spectral databases. The identification is severely limited by the available spectra in the databases. Although, the metabolome consists of a huge number of different functional metabolites, the whole metabolome derives from a limited number of initial metabolites via bioreactions. In each bioreaction, the reactant and the product often change some substructures but are still structurally related. These structurally related metabolites often have related MS/MS spectra, which provide the possibility to identify unknown metabolites through known ones. However, it is challenging to explore the internal relationship between MS/MS spectra and structural similarity. In this study, we present the deep-learning-based approach for MS/MS-aided structural-similarity scoring (DeepMASS), which can score the structural similarity of unknown metabolite against the known one with MS/MS spectra and deep neural networks. We evaluated DeepMASS with leave-one-out cross-validation on MS/MS spectra of 662 compounds in KEGG and an external test on the biomarkers from male infertility study measured on Shimadzu LC-ESI-IT-TOF and Bruker Compact LC-ESI-QTOF. Results show that the identification of unknown compound is valid if its structure-related metabolite is available in the database. It provides an effective approach to extend the identification range of metabolites for existing MS/MS databases.

No abstract

Distilling accurate quantitation information on metabolites from liquid chromatography coupled with mass spectrometry (LC-MS) data sets is crucial for further statistical analysis and biomarker identification. However, it is still challenging due to the complexity of biological systems. The concept of pure ion chromatograms (PICs) is an effective way of extracting meaningful ions, but few toolboxes provide a full processing workflow for LC-MS data sets based on PICs. In this study, an integrated framework, KPIC2, has been developed for metabolomics studies, which can detect pure ions accurately, align PICs across samples, group PICs to identify isotope and potential adducts, fill missing peaks and do multivariate pattern recognition. To evaluate its performance, MM48, metabolomics quantitation, and Soybean seeds data sets have been analyzed using KPIC2, XCMS, and MZmine2. KPIC2 can extract more true ions with fewer detecting features, have good quantification ability on a metabolomics quantitation data set, and achieve satisfactory classification on a soybean seeds data set through kernel-based OPLS-DA and random forest. It is implemented in R programming language, and the software, user guide, as well as example scripts and data sets are available as an open source package at https://github.com/hcji/KPIC2 .

No abstract

Electron−ionization mass spectrometry (EI-MS) hyphenated gas chromatography (GC) is the workhorse to analyze volatile compounds in complex samples. The spectral matching method can only identify compounds within spectral database. In response, we present a deep-learning-based approach (DeepEI) for structure elucidation of unknown compound with its EI-MS spectrum. DeepEI employs deep neural networks to predict molecular fingerprint from EI-MS spectrum, and searches molecular structure database with the predicted fingerprints. In addition, a convolutional neural network was also trained to filter the structures in database and improve the identification performance. Our method shows improvement on the competing method NEIMS in identification accuracy on both NIST test dataset and MassBank dataset. Furthermore, DeepEI (spectrum to fingerprint) and NEIMS (fingerprint to spectrum) can be combined to improve identification accuracy.