NPOmix: A machine learning classifier to connect mass spectrometry fragmentation data to biosynthetic gene clusters

Tiago F Leão; Mingxun Wang; Ricardo da Silva; Alexey Gurevich; Anelize Bauermeister; Paulo Wender P Gomes; Asker Brejnrod; Evgenia Glukhov; Allegra T Aron; Joris J R Louwen; Hyun Woo Kim; Raphael Reher; Marli F Fiore; Justin J J van der Hooft; Lena Gerwick; William H Gerwick; Nuno Bandeira; Pieter C Dorrestein

doi:10.1093/pnasnexus/pgac257

NPOmix: A machine learning classifier to connect mass spectrometry fragmentation data to biosynthetic gene clusters

PNAS Nexus. 2022 Nov 16;1(5):pgac257. doi: 10.1093/pnasnexus/pgac257. eCollection 2022 Nov.

Authors

Tiago F Leão^{1

2}, Mingxun Wang^{1

3}, Ricardo da Silva⁴, Alexey Gurevich⁵, Anelize Bauermeister¹, Paulo Wender P Gomes¹, Asker Brejnrod¹, Evgenia Glukhov⁶, Allegra T Aron^{1

7}, Joris J R Louwen⁸, Hyun Woo Kim⁹, Raphael Reher¹⁰, Marli F Fiore², Justin J J van der Hooft^{8

11}, Lena Gerwick⁶, William H Gerwick^{1

6}, Nuno Bandeira^{1

3}, Pieter C Dorrestein^{1

12

13}

Affiliations

¹ Collaborative Mass Spectrometry Innovation Center, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California San Diego, La Jolla, CA 92093, USA.
² Center for Nuclear Energy in Agriculture, University of São Paulo, Piracicaba 13400-970, SP, Brazil.
³ Center for Computational Mass Spectrometry, University of California San Diego, La Jolla, CA 92093, USA.
⁴ NPPNS, Physic and Chemistry Department, School of Pharmaceutical Sciences of Ribeirão Preto, University of São Paulo, Ribeirão Preto 14040-900, Brazil.
⁵ Center for Algorithmic Biotechnology, St. Petersburg State University, St Petersburg 199004, Russia.
⁶ Center for Marine Biotechnology and Biomedicine, Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA 92093, USA.
⁷ Department of Chemistry and Biochemistry, University of Denver, Denver, CO 80210, USA.
⁸ Bioinformatics Group, Wageningen University, 6708 PB Wageningen, The Netherlands.
⁹ College of Pharmacy and Integrated Research Institute for Drug Development, Dongguk University, Gyeonggi-do 10326, Korea.
¹⁰ Institute of Pharmaceutical Biology and Biotechnology, University of Marburg, 35043 Marburg, Germany.
¹¹ Department of Biochemistry, University of Johannesburg, Auckland Park, Johannesburg 2006, South Africa.
¹² Center for Microbiome Innovation, University of California San Diego, La Jolla, CA 92093, USA.
¹³ Departments of Pharmacology and Pediatrics, University of California San Diego, La Jolla, CA 92093, USA.

Abstract

Microbial specialized metabolites are an important source of and inspiration for many pharmaceuticals, biotechnological products and play key roles in ecological processes. Untargeted metabolomics using liquid chromatography coupled with tandem mass spectrometry is an efficient technique to access metabolites from fractions and even environmental crude extracts. Nevertheless, metabolomics is limited in predicting structures or bioactivities for cryptic metabolites. Efficiently linking the biosynthetic potential inferred from (meta)genomics to the specialized metabolome would accelerate drug discovery programs by allowing metabolomics to make use of genetic predictions. Here, we present a k-nearest neighbor classifier to systematically connect mass spectrometry fragmentation spectra to their corresponding biosynthetic gene clusters (independent of their chemical class). Our new pattern-based genome mining pipeline links biosynthetic genes to metabolites that they encode for, as detected via mass spectrometry from bacterial cultures or environmental microbiomes. Using paired datasets that include validated genes-mass spectral links from the Paired Omics Data Platform, we demonstrate this approach by automatically linking 18 previously known mass spectra (17 for which the biosynthesis gene clusters can be found at the MIBiG database plus palmyramide A) to their corresponding previously experimentally validated biosynthetic genes (e.g., via nuclear magnetic resonance or genetic engineering). We illustrated a computational example of how to use our Natural Products Mixed Omics (NPOmix) tool for siderophore mining that can be reproduced by the users. We conclude that NPOmix minimizes the need for culturing (it worked well on microbiomes) and facilitates specialized metabolite prioritization based on integrative omics mining.

Keywords: biosynthetic gene clusters; genomics; machine learning; mass spectrometry; specialized metabolites.

Grants and funding

R01 GM118815/GM/NIGMS NIH HHS/United States