Supervised topic modeling for predicting molecular substructure from mass spectrometry

Abstract

Small-molecule metabolites are principal actors in myriad phenomena across biochemistry and serve as an important source of biomarkers and drug candidates. Given a sample of unknown composition, identifying the metabolites present is difficult given the large number of small molecules both known and yet to be discovered. Even for biofluids such as human blood, building reliable ways of identifying biomarkers is challenging. A workhorse method for characterizing individual molecules in such untargeted metabolomics studies is tandem mass spectrometry (MS/MS). MS/MS spectra provide rich information about chemical composition. However, structural characterization from spectra corresponding to unknown molecules remains a bottleneck in metabolomics. Current methods often rely on matching to pre-existing databases in one form or another. Here we develop a preprocessing scheme and supervised topic modeling approach to identify modular groups of spectrum fragments and neutral losses corresponding to chemical substructures using labeled latent Dirichlet allocation (LLDA) to map spectrum features to known chemical structures. These structures appear in new unknown spectra and can be predicted. We find that LLDA is an interpretable and reliable method for structure prediction from MS/MS spectra. Specifically, the LLDA approach has the following advantages: (a) molecular topics are interpretable; (b) A practitioner can select any set of chemical structure labels relevant to their problem; (c ) LLDA performs well and can exceed the performance of other methods in predicting substructures in novel contexts.

Multiomics characterization of cell type repertoires for urine liquid biopsies

“…We observed metabolites enriched in the same pathways we observed in the urinary transcriptome, and we further observed that metabolic cell types whose signal we measure with RNA participate in distinct metabolic pathways with measurable metabolites. We are optimistic that the ability to use metabolomics as a functional readout alongside gene expression changes in linking genotype to disease phenotype and enactment will improve with the development of new tools that expand the space of identifiable compounds from untargeted metabolomics data 49,50 and the continued joint measurement of the transcriptome and metabolome, thereby offering increased potential for more precise biomarker panels for disease diagnostics and subtyping. This work also identifies and addresses challenges facing the development of urine liquid biopsies.…”

A mathematical comparison of non‐negative matrix factorization related methods with practical implications for the analysis of mass spectrometry imaging data

“…For MALDI‐TOF MSI, it has been used as a soft clustering technique and, with tandem mass spectrometry (MS/MS), one can use LDA to cluster the fragmentation of the data in substructures, as has been done in previous research. 33 , 34 , 35 , 36 …”

“…For MALDI-TOF MSI, it has been used as a soft clustering technique and, with tandem mass spectrometry (MS/MS), one can use LDA to cluster the fragmentation of the data in substructures, as has been done in previous research. [33][34][35][36] Let us define similarly as for PLSA: N as the number of m/z bins (or words) in the data, K as the number of components (or topics), which can be seen as the different cell-types within the tissue, and D as the number of spectra or pixels in the MSI dataset (which corresponds to the number of documents in the setting of text mining). Also, let θ d be the probability of a spectrum (d) belonging to a cell type (c) according to a Dirichlet prior α, and β c be the probability of a certain m/z bin (w) belonging to a cell type (c) according to another Dirichlet prior η.…”

Recent advances in data- and knowledge-driven approaches to explore primary microbial metabolism

“…Reder et al. developed Labeled Latent Dirichlet Allocation to map spectrum features to the chemical space of known structures as a supervised topic modeling approach, which allows for interpretable chemical structure prediction given tandem MS profiles [ 41 ]. Gao et al.…”