PeakDecoder enables machine learning-based metabolite annotation and accurate profiling in multidimensional mass spectrometry measurements - Research overview

Aug 5, 2026

Upload by Aurora

Multidimensional measurements integrating advanced separations and mass spectrometry improve untargeted metabolomics for biological and environmental biochemical processes. Limited application has stemmed from the absence of rapid analytical methods and robust algorithms for heterogeneous data. The work presents a sensitive, high-throughput experimental and computational workflow that combines LC, ion mobility spectrometry, and data-independent acquisition MS with PeakDecoder, enabling accurate metabolite profiling and estimation of identification error rates across diverse engineered microbial strains.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages38

Reading Time1 h 36 min

AccessFree to download

Abstract

Introduction

Methods and workflow

PeakDecoder algorithm for identification error rates

Experimental validation and results

Application to engineered microbial strains

Lawrence Berkeley National Laboratory Biological Systems & Engineering Title PeakDecoder enables machine learning-based metabolite annotation and accurate profiling in multidimensional mass spectrometry measurements Permalink [https://escholarship.org/uc/item/6mx6n2bc](https://escholarship.org/uc/item/6mx6n2bc) Journal Nature Communications, 14(1) ISSN 2041-1723 Authors Bilbao, Aivett Munoz, Nathalie Kim, Joonhoon et al. Publication Date 2023 DOI 10.1038/s41467-023-37031-9 Copyright Information This work is made available under the terms of a Creative Commons Attribution-ShareAlike License, availalbe at [https://creativecommons.org/licenses/by-sa/4.0/](https://creativecommons.org/licenses/by-sa/4.0/) Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California 1 PeakDecoder enables machine learning-based metabolite annotation 2 and accurate profiling in multidimensional mass spectrometry 3 measurements 4 5 Aivett Bilbao 1,2* , Nathalie Munoz 1,2 , Joonhoon Kim1,2 , Daniel J Orton 1 , Yuqian Gao 1,2 , Kunal 6 Poorey3 , Kyle R. Pomraning 1,2 , Karl Weitz 1 , Meagan Burnet 1 , Carrie D. Nicora 1 , Rosemarie 7 Wilton4,2 , Shuang Deng 1,2 , Ziyu Dai 1,2 , Ethan Oksen5 , Aaron Gee6 , Rick A. Fasani6 , Anya 8 Tsalenko6 , Deepti Tanjore5,2 , James Gardner5,2 , Richard D. Smith 1 , Joshua K. Michener7,2 , John 9 M. Gladden3,2 , Erin S. Baker8 , Christopher J. Petzold5,2 , Young-Mo Kim 1,2 , Alex Apffel6 , Jon K. 10 Magnuson 1,2 and Kristin E. Burnum-Johnson 1,2* 11 12 1 Pacific Northwest National Laboratory, Richland, WA, USA 13 2 US Department of Energy, Agile BioFoundry, Emeryville, CA, USA 14 3 Sandia National Laboratory, Livermore, CA, USA 15 4 Argonne National Laboratory, Lemont, IL, USA 16 5 Lawrence Berkeley National Laboratory, Berkeley, CA, USA 17 6 Agilent Research Laboratories, Agilent Technologies, Santa Clara, CA, USA 18 7 Oak Ridge National Laboratory, Oak Ridge, TN, USA 19 8 Department of Chemistry, University of North Carolina, Chapel Hill, NC, USA 20 21 These authors contributed equally: Aivett Bilbao, Nathalie Munoz, and Joonhoon Kim. 22 23 * Correspondence: 24 [Aivett.Bilbao@pnnl.gov](Aivett.Bilbao@pnnl.gov) and [Kristin.Burnum-Johnson@pnnl.gov](Kristin.Burnum-Johnson@pnnl.gov) 25 Abstract 26 Multidimensional measurements using state-of-the-art separations and mass spectrometry 27 provide advantages in untargeted metabolomics analyses for studying biological and 28 environmental bio-chemical processes. However, the lack of rapid analytical methods and robust 29 algorithms for these heterogeneous data has limited its application. Here, we develop and 30 evaluate a sensitive and high-throughput analytical and computational workflow to enable 31 accurate metabolite profiling. Our workflow combines liquid chromatography, ion mobility 32 spectrometry and data-independent acquisition mass spectrometry with PeakDecoder, a machine 33 learning-based algorithm that learns to distinguish true co-elution and co-mobility from raw data 34 and calculates metabolite identification error rates. We apply PeakDecoder for metabolite profiling 35 of various engineered strains of Aspergillus pseudoterreus, Aspergillus niger, Pseudomonas 36 putida and Rhodosporidium toruloides. Results, validated manually and against selected reaction 37 monitoring and gas-chromatography platforms, show that 2683 features could be confidently 38 annotated and quantified across 116 microbial sample runs using a library built from 64 standards. 39 40 Introduction 41 Metabolomics is the study of the small molecules produced by complex networks of cellular 42 processes and biochemical reactions in living systems. Metabolites are the end point of the flow 43 of information from DNA to the biological phenotype and represent chemical fingerprints directly 44 reflecting the physiological co

FAQs

What problem does PeakDecoder address in untargeted metabolomics?

It targets the lack of rapid analytical methods and robust algorithms to handle heterogeneous, multidimensional mass spectrometry data for accurate metabolite identification and profiling.

What experimental workflow is combined with PeakDecoder?

The workflow integrates liquid chromatography, ion mobility spectrometry, and data-independent acquisition mass spectrometry, using PeakDecoder to learn true co-elution and co-mobility patterns from raw data.

How are metabolite identification results validated?

Results are validated manually and compared against selected reaction monitoring and gas-chromatography platforms.

Aug 5, 2026

Upload by Aurora