Annotating metabolite mass spectra with domain-inspired chemical formula transformers

Abstract

Metabolomic studies have succeeded in identifying small molecule metabolites that mediate cell signaling, competition, and disease pathology in part due to large-scale community efforts to measure mass spectra for thousands of metabolite standards. Nevertheless, the vast majority of spectra observed in clinical samples cannot be unambiguously matched to known structures, suggesting powerful opportunities for further discoveries in the dark metabolome. Deep learning approaches to small molecule structure elucidation have surprisingly failed to rival classical statistical methods, which we hypothesize is due to the lack of in-domain knowledge incorporated into current neural network architectures. We introduce a new neural network driven workflow for untargeted metabolomics, Metabolite Inference with Spectrum Transformers (MIST), to annotate mass spectrometry peaks with chemical structures generalizing beyond known standards. Unlike other neural approaches, MIST incorporates domain insights into its architecture by forcing the network to more directly link peaks to physical atom representations, neutral losses, and chemical substructures. MIST outperforms both standard neural architectures and the state-of-the-art kernel method on fingerprint prediction from spectra for over 70% of metabolite standards and retrieves over 66% of metabolites with equal or improved accuracy, with 29% strictly better. We further demonstrate the utility of MIST in a prospective setting to identify new differentially abundant metabolite structures from an inflammatory bowel disease patient cohort and subsequently annotate dipeptides and alkaloid compounds without spectral standards.

Refining EI-MS library search results through atomic-level insights

“…The deduction of key aspects of a given compound such as molecular formula, structural features like side groups or substructures, and overall molecular structure, heavily relies on the accurate annotation of mass spectra [35,36]. Incorrect annotations are likely to amplify structural dissimilarities, prompting a reevaluation and correction of the molecular formula annotations.…”

Connecting metabolome and phenotype: recent advances in functional metabolomics tools for the identification of bioactive natural products

“…To bypass the limitations of MS/MS spectral libraries, an alternative computational approach translates MS/MS information directly into structural information as structural ngerprints. 163 CSI:FingerID 164 and MIST 165 utilize machine learning to predict these ngerprints directly from the MS/MS data. CANOPUS 166 utilizes these predicted ngerprints, which are esoteric to chemists, to predict the natural product class 167 of the unknown natural product.…”

MS2Mol: A transformer model for illuminating dark chemical space from mass spectra

“…Examples of these methods are CSI:FingerID [17], implemented in the SIRIUS software suite, and the tool used in the highest-accuracy submission to the recent Critical Assessment of Small Molecule Identification (CASMI) contest [18]. This is also used by MIST [19], which predicts the same CSI:FingerID fingerprints using a neural network.…”