Skip to main content
Nucleic Acids Research logoLink to Nucleic Acids Research
. 2025 Nov 22;54(D1):D623–D634. doi: 10.1093/nar/gkaf1272

MiMeDB 2.0: the Human Microbial Metabolome Database for 2026

Ray Kruger 1, Eponine Oler 2, Sukanta Saha 3,4, Jenna Poelzer 5, Scott Han 6, Mark Peter Punsalan 7, Brittany Green 8, Fahrin Bushra 9, Megane Kyes 10, Fehintola Disu 11, Julia Wakoli 12, Robyn Woudstra 13, Elykah Tejol 14, Omolola Fakutan 15, Fei Wang 16, Brian L Lee 17, Tanvir Sajed 18, Siyang Tian 19, Claudia Torres-Calzada 20, Mark Berjanskii 21, Scott MacKay 22, Naama Karu 23, Rima Kadurrah-Daouk 24, David S Wishart 25,26,27,28,✉
PMCID: PMC12807671  PMID: 41273085

Abstract

The Microbial Metabolome Database (MiMeDB) (https://mimedb.org) is a comprehensive, freely accessible resource linking human-associated microbes to the metabolites they produce, along with their connections to human health, disease, and diet. Since the release of MiMeDB 1.0 in 2023, the database has been substantially expanded and redesigned. Major updates include the systematic addition of millions of newly annotated genes and pathways, thousands of new metabolites, significantly expanded pathway and reaction coverage, along with broader representation of eukaryotic gut microbes. MiMeDB 2.0 now contains >12.9 million annotated microbial genes, over 23.1 million microbial pathways, 29 295 metabolites, 21 829 metabolic reactions, 3725 microbial species and strains, and 514 076 new experimental and predicted nuclear magnetic resonance and mass spectrometry spectra of microbial metabolites. New features, such as detailed microbial descriptions, metabolite origin tags, refined search filters, and species-specific reaction queries, have been added to enhance usability. Likewise, redesigned network and genome viewers have been implemented to support more comprehensive, intuitive, and integrated visualization of complex, multi-omic relationships. The significant addition of more metabolite spectral data and improved spectral search capabilities further strengthen metabolite identification and discovery. Together, these improvements make MiMeDB 2.0 one of the most comprehensive and user-friendly platforms for investigating the human microbiome at a molecular level and exploring the roles of microbes and microbial metabolites in human health, diet, and disease.

Graphical Abstract

Graphical Abstract.

Graphical Abstract

Introduction

Over the past two decades, the study of the human microbiome has advanced at an extraordinary pace, driven by breakthroughs in genomics, proteomics, and metabolomics. This research has revealed that human-associated microbes influence health and disease not merely through their biological presence, but largely through the small-molecule metabolites they produce [1]. Many of these compounds are now known to influence human physiology in strikingly different ways: from protective molecules such as butyrate, which reduce inflammation and cancer risk [2], to harmful metabolites, such as indoxyl sulfate, which is associated with cardiovascular toxicity [3] and cognitive impairment [4]. As a result, the systematic study of microbial metabolites has become central to understanding host–microbe interactions and the biochemical basis of microbiome-related health and diet outcomes.

Despite this growing recognition, comprehensive information on microbial metabolites, along with information on microbial metabolic pathways, genomes, proteomes, and metabolomes, has remained fragmented, disconnected, and widely scattered. Microbial metabolic pathway data can be found in multiple resources such as the Kyoto Encyclopedia of Genes and Genomes (KEGG) [5], BioCyc [6], and the Virtual Metabolic Human (VMH) [7]; microbial genomic and proteomic data reside in databases such as GenBank [8], the Unified Human Gastrointestinal Genome (UHGG) [9], and UniProt [10]; microbial chromosome maps are housed in BacMap [11] and Ensembl Bacteria [12]; while resources such as the Human Metabolome Database (HMDB) [13], the Escherichia coli Metabolome Database [14], the Yeast Metabolome Database (YMDB) [15], MarkerDB [16], and FooDB (https://foodb.ca) provide data on microbial metabolite descriptions, dietary exposures, or microbe-associated health effect annotations. However, none of these databases were explicitly developed to unify and integrate microbial multi-omics data within the context of the human microbiome or to capture their broader metabolic implications. Among the databases that do focus on microbial information, Microbial Metabolome Database (MiMeDB) stands out as the most comprehensive, offering data on 29 295 microbial metabolites—a collection that significantly exceeds the size of other available resources. For comparison, gutMGene (version 2.0) contains data on 830 microbial metabolites, KEGG (release 116) includes ~7700, BioCyc (version 29.1) has 3794, BacDive (2023.2 release) [17] catalogs 801, and the VMH database (as of May 2019) includes 5607. MiMeDB also offers many additional advantages over these databases (see Table 1).

Table 1.

Comparison of MiMeDB 2.0 to other available Microbial Metabolite Databases

Databases MiMeDB 2.0 KEGG GutMGene BioCyc BacDive VMH
# of microbes 3725 10 207 4744 19 448 99 393 818
# of metabolites 29 295 7 700 830 3794 767 5607
Pathway diagrams Yes Yes Yes Yes Yes Yes
Genome sequenceCoverage Human associated bacteria & fungi All microbes No genome data All microbes No genome data No genome data
# of individual microbial genes 12 326 180 40 209 708 Host genes only 65 291 936 41 458 632 685
Biocard (microbe description) information Yes No No Yes (limited) Yes Yes (limited)
Visualization tools Network Viewer, Genome Viewer Network Maps No Genome Explorer No Leigh Maps, Recon Maps, Numerical Characteristics Graphs
ReactionDiagrams Yes Yes Yes Yes No Yes
Protein (enzyme)Information Yes Yes No Yes Yes Yes
Exposure sources Yes No No No Yes (count only) Yes
Associated health conditions Yes Yes (disease pathways) Yes No Yes (limited) Yes

To address this gap, we launched the MiMeDB in 2023. MiMeDB 1.0 provided the first freely accessible, fully integrated resource dedicated to the integration of human microbiome data with microbial metabolome, proteome, genome, and exposome data. The goal of MiMeDB was to facilitate more detailed interpretation and understanding of the interactions between the microbiome and human health, human metabolism, and human nutrition. However, as with any initial release, MiMeDB 1.0 had several limitations. Key metabolite classes, such as secondary bile acids and certain microbial lipids, were underrepresented; many microbe-specific pathways were absent; genome/proteome annotations were incomplete; and many newly discovered microbes were absent. Users also noted functional limitations, including the lack of filtered data downloads, a cluttered network viewer, and a genome browser that was not fully developed nor capable of viewing eukaryotic/fungal genomes.

In response to these shortcomings and in response to extensive feedback from the community, we have developed MiMeDB 2.0, a substantially expanded and redesigned version of the original database. This significantly updated release incorporates hundreds of additional microbes, thousands of new metabolites, as well as thousands of new health effect relations and metabolic reactions. Equally importantly, MiMeDB 2.0 includes millions more pathways and richly annotated genes and proteins, along with thousands of additional nuclear magnetic resonance (NMR), mass spectrometry (MS), and chromatographic data to aid in microbial/fecal/urinary metabolite identification. Genome maps generated with tools such as BASys2 [18] and EUKLID (https://euklidgenome.ca) have been added, enhancing the depth, breadth, and viewability of genomic/proteomic data. The MiMeDB Network Viewer has been rebuilt for greater clarity and usability, and system optimizations have significantly improved performance and loading times. Collectively, these improvements make MiMeDB 2.0 a much more comprehensive, integrated, and versatile resource for exploring the multi-omic links between microbes, metabolism, diet, health and disease.

Expanded database content

Building on the extensive foundation built for MiMeDB 1.0, our team has systematically expanded MiMeDB’s content, adding thousands (even millions) of new entries across microbial species, metabolites, genes, proteins, pathways, reactions, and spectra. The new release includes 1821 new microbes, 5041 additional metabolites, 11 584 new metabolic reactions, 9.4 million newly annotated genes, 12.4 million new proteins, 23.1 million newly mapped pathways, and 514 076 newly measured or accurately predicted MS (including MS/MS) and NMR spectra (Table 2). Incorporating these data required the development of several new annotation tools and annotation pipelines [19–24], the refinement of many other software and data visualization programs [25–27], extensive manual literature reviews, careful reformatting, and multiple layers of manual vetting by the MiMeDB curation team.

Table 2.

Overview of updates in metabolites, microbes, associated health effects, reactions, pathways, and spectral data from MiMeDB 1.0 to MiMeDB 2.0

Features MiMeDB 1.0 MiMeDB 2.0 Number increase % Increase
No. of metabolites 24 254 29 295 5041 20.8
No. of microbes 1904 3725 1821 95.6
No. of bacteria 1896 3287 1391 73.4
No. of bacterial genomes 1896 3225 1329 70.1
No. of bacterial genes annotated 3 487 109 12 326 180 8 839 071 253.5
No. of bacterial proteins 197 094 12 046 958 11 849 864 6012.3
No. of fungi 8 325 317 3962.5
No. of fungal genomes 8 85 77 962.5
No. of fungal genes annotated 9867 549 049 539 182 5 464.5
No. of fungal proteins − 525 670 525 670 NA
No. of biospecimen types 18 55 37 205.6
No. of health condition relationships 641 7379 6738 1051.2
No. of metabolic reactions 13 692 25 276 11 584 84.6
No. of metabolites with pathways 19 389 25 941 6552 33.8
No. of metabolic and signaling pathways 25 941 23 106 015 23 080 074 88 971.4
No. of exposure sources 767 5901 5134 669.4
No. of references 3112 23 379 20 267 651.3
No. of chemical classes 72 211 139 193.1
No. of health conditions 626 1811 1185 189.3
No. of microbe related health effects 3990 9676 5686 142.5
No. of biospecimen/metabolite relations 3066 8681 5615 183.1
No. of human/microbial co-metabolites 50 280 230 460.0
No. of microbial-only metabolites 1808 4932 3124 172.8
No. of metabolites common to human + microbes 14 210 21 705 7495 52.7
No. of exogenously sourced metabolites 1524 2378 854 56.0
No. of predicted retention time values – 537 416 537 416 NA
No. of predicted retention index values – 190 773 190 773 NA
No. of predicted 1H NMR 1D spectra 61 294 270 111 208 817 308.1
No. of predicted 13C NMR 1D spectra 6128 27 022 20 894 341
No. of predicted 2D NMR spectra – 27 022 27 022 NA
No. of predicted MS/MS spectra 92944 278 268 185 324 199.4
No. of experimental MS/MS spectra 20 784 35 404 14 620 70.3
No. of predicted EI-MS spectra 37 024 63 591 26 567 71.8
No. of experimental EI-MS spectra 5058 7402 2344 46.3
No. of predicted IR spectra – 2423 2423 NA
No. of experimental IR spectra – 60 60 NA

To ensure reliability of the manually entered data (specifically for microbial species, metabolites, biospecimens, pathways/reactions, associated health effects, and diet/exposure data), only peer-reviewed sources were included. Strong preference was given to studies providing precise microbial data (species and strain ID), detailed metabolite data, experimentally validated pathways, and experimentally confirmed chemical reactions or products. Special focus was also given to published studies that had clear nutritional or clinical relevance in humans. Exclusion criteria included non-peer-reviewed reports, animal-only studies without demonstrated human analogs, and in vitro findings lacking validation. The following subsections outline these expansions in detail, with a focus on newly added microbial species, microbial metabolites, genome/proteome data, and pathways.

Expanded coverage of microbial species

A key focus in upgrading MiMeDB was the expansion of microbial species contained in the database, both newly discovered organisms and those omitted from the initial release. These new microbes were identified and then gathered from the latest releases of VMH (release May 2019) [7], BacDive (release 2023.2) [17], the NCBI Refseq database (release 232) [28], scientific articles researched by our curation team, and the UHGG (version 2.0.2) [9]. This led to the creation of a microbial organism list that had 3287 bacterial species and 328 additional fungal species. This brings the total number of microbes in MiMeDB 2.0 to 3347, which includes 3251 eubacteria, 59 archaebacteria, 22 viruses, and 325 fungi, which is 95.6% more than MiMeDB 1.0 (see Table 2). To improve consistency and clarity, microbial data have been consolidated at the species level, with strain- or variant-specific information displayed on each species page (Fig. 1). This approach avoids duplication and ensures that all variant-level details are unified under a single, comprehensive species entry. Notably, 41 species that were previously represented only by individual variants are now included as full species entries. In addition to improving the taxonomic integrity of the database, each bacterial entry was systematically updated to include comprehensive biological and physiological traits, such as gram-staining properties, oxygen requirements, energy sources, cell shape, motility, flagella presence, sporulation status, cell arrangement, and biotic relationships.

Figure 1.

Figure 1.

Screenshots from the MiMeDB 2.0 microbe page showing (A) microbe identification with the new description and linked species, and (B) the pop-up window that appears when a species is clicked, here displaying variants of Acinetobacter baumannii.

A significant enhancement to MiMeDB 2.0 has been the expanded representation of eukaryotic microbes, particularly intestinal fungi. In this update, 317 fungal species were added, bringing the total number of fungi in the database to 325. Similar to the bacterial annotations, additional information was provided on fungal physiological and biological traits such as oxygen requirements, energy sources, sporulation status, cell arrangement, and biotic relationship. To facilitate the collection of this trait information, a hybrid curation strategy was implemented. A large language model (LLM) (https://openai.com/) was employed to initially identify candidate research articles and extract potential trait data. This was then used to parse and summarize data from articles containing the microbe’s name, then return information for each of the categories as described earlier. Because automated outputs can contain errors, every entry was manually reviewed by the curation team. In cases where data were incomplete or inconsistent, manual literature searches were conducted to supplement and verify the information. This combined approach accelerated the data-gathering process while maintaining accuracy and reliability.

In addition to the expanded number of microbes and the inclusion of detailed biological trait data, MiMeDB 2.0 now provides detailed descriptions (ranging from 150 to 300 words) for all 2550 microbial entries currently in the database. These summaries were generated using a locally developed retrieval augmented generation LLM based on ChatGPT 4o mini, which synthesized curated database content with information retrieved from MiMeDB’s tabular data and external references. Each summary was subsequently manually reviewed and edited by the curation team. These information-rich overviews provide accessible descriptions of microbial characteristics, enhancing the usability of MiMeDB for both researchers and non-specialist audiences.

Expanded metabolome coverage

To expand MiMeDB 2.0’s metabolite coverage, a systematic, manually intensive search strategy was employed across PubMed, Scopus, and the Web of Science, using combinations of MeSH terms and keywords such as “metabolome,” “human,” “microbes,” and “microbial metabolites” combined with Boolean operators (e.g. “metabolome AND microbes,” “human AND,” and “microbial metabolites”). This manual data extraction process was supported by several in-house tools for literature mining, metabolite annotation, and format standardization (Chemosummarizer and Data Wrangler) [12], followed by manual review to confirm accuracy and consistency. Each record was then curated to include metabolite structures, metabolite descriptions, concentrations (where available), associated microbes, metabolite origin(s), pathways, and links to health outcomes or dietary intake. This rigorous selection and curation process ensured that MiMeDB 2.0 provided substantially broader and higher-quality microbial metabolite data than its predecessor. Indeed, compared with MiMeDB 1.0, the number of metabolites in MiMeDB 2.0 increased from 24 254 to 29 295, representing a 20.8% growth (see Table 2).

A central focus in expanding and enriching MiMeDB 2.0’s metabolite coverage was aimed at expanding the identification and annotation of important microbial and human–microbial co-metabolites (such as indoxyl and indoxyl sulfate). A microbial/host co-metabolite is any metabolite that requires consecutive biochemical reactions from the host and gut microbiota, not in that particular order. One option is a co-metabolite that originates in a human endogenous metabolite that undergoes microbial metabolism, e.g. secondary bile acids (from primary) and urobilins (from red blood cells breakdown into heme and bilirubin). The second option is a gut microbial product that is absorbed and then undergoes human metabolism (typically phase I and II biotransformation), such as TMAO, hippuric acid, p-cresol sulfate and indoxyl glucuronide. This process usually involves ingested dietary substrates, such as essential amino acids, plant phenolic compounds, and animal products, which are also human endogenous (carnitine, choline etc.). These compounds represent the most direct biochemical interface between the microbiome and the human host. Particular emphasis was placed on finding and identifying co-metabolites, with 100 new entries systematically added. These were sourced primarily from the Collaborative Microbial Metabolite Center Knowledgebase or CMMC-kb (https://cmmc.gnps2.org), a community-contributed repository of microbial metabolites, as well as the UHGG [9], and through additional manual curation from the literature. Notably, many of these additions include 181 secondary bile acids, a prominent class of co-metabolites with well-documented roles in gut physiology, signaling, and disease pathogenesis, and 696 acylamides. Additional emphasis was placed on adding more microbial-specific lipids to MiMeDB. Microbes produce a wide range of unusual lipids (such as lipopolysaccharides [LPSs], lipotechoic acids [LTAs], and cyclic Caldarchaeol lipids), which add considerable variability to both the human and microbial lipidome. Many of these lipid structures were sourced from the Pathways for Bacteria Knowledgebase (PathBacK) (https://pathback.wishartlab.com/), a knowledgebase of bacterial pathways that compiles experimentally supported lipid data and lipid reactions from hundreds of studies alongside over 35 000 predicted lipid reactions through the predicted pathways.

To more clearly represent metabolite origin, an essential feature for microbial metabolite characterization, origin tags have been added to the top of each MiMeDB MetaboCard. These tags allow users to immediately identify whether a metabolite is classified as purely microbial or a co-metabolite, or whether it arises from human, animal, plant, pharmaceutical, or “other” xenobiotic origins. While detailed origin information remains accessible through MiMeDB’s exposure sources and reaction sections, the new tags highlight key classifications at a glance. Metabolite origin was determined by manually analyzing the enzymatic reactions in which each compound appears as a product within MiMeDB. These reactions were curated from literature as well as KEGG [5], VMH [7], RHEA [27], and PathBank [25], with additional annotations derived from recently updated PathBank data. This approach enabled a more accurate distinction between microbial, human, and co-metabolite origins, further enriching the metabolite information available in MiMeDB.

Expanded genome and proteome coverage

The past 3 years have seen a significant expansion in the number and quality of microbial reference genomes housed at the NCBI. Now nearly all known human microbial species and strains have fully sequenced genomes. This provided the MiMeDB curation team with the opportunity to significantly expand the depth and breadth of its genome and proteome annotations. In particular, the MiMeDB team used two recently developed tools, BASys2 [18] and EUKLID (https://euklidgenome.ca), to perform in-depth genome and proteome annotations for nearly all microbes in MiMeDB 2.0. Genomes were selected and downloaded based on the organism list (described earlier) that had both species and/or strain and/or subspecies information. Name comparisons, combined with sequence content, genome length, and genome sequence analysis, were used to confirm matches to the NCBI’s Refseq [28] data. Only those microbes with completed genomes in Refseq were used to compile the final MiMeDB 2.0 genome set. Proteome data were determined via BASys 2.0 or data in Refseq. In total, >12.6 million protein-coding genes and 302 000 non-coding genes were annotated across all microbes in MiMeDB 2.0. This represents an expansion of >268% in terms of gene numbers and >149% in terms of the number of genome/proteome annotations (see Table 2).

Both BASys2 and EUKLID employ >30 different bioinformatics programs and 10 different databases to assemble and/or annotate any given prokaryotic or eukaryotic genome, yielding up to 62 annotation fields for each gene. These data fields cover both genomic and proteomic information and include detailed gene position data (start, end, exon/intron boundaries), gene sequence data, transfer RNA, messenger RNA, soluble RNA, ribosomal RNA sequence data, protein sequence data, operon structures (for bacteria), prophage sequences (for bacteria), protein functional data (COG [29] and GO [30]), reaction data (from RHEA [27] and PathBacK), signal sequence data (for both bacteria and fungi), biophysical data (cell location, sequence length, isoelectric point, molecular weight, average hydrophobicity, selected amino acid content, membrane spanning regions), secondary structure, 3D structure (both experimental and predicted via AlphaFold2 [31]), and in-depth metabolite/metabolome annotations. This greatly enriched genome/proteome data complements the metabolome data and enabled MiMeDB to provide vastly improved multi-omics integration and opened the door to significantly expanded pathway coverage.

Expanded pathway and reaction coverage

In addition to expanding the metabolome, genome, and proteome coverage, MiMeDB 2.0 greatly enhanced its pathway and reaction coverage. Incorporation of PathBacK data has enabled MiMeDB 2.0 to link 25 941 microbial metabolites directly to known enzymatic reactions and pathways, providing contextual information that extends beyond metabolite identities alone. This pathway-centric annotation is not widely available in other microbial or other metabolite databases, and we believe it is key to capturing the mechanistic basis of metabolite production in microbes.

In total, >23.1 million pathways were added to MiMeDB 2.0. This represents a 716-fold expansion over the number of pathways in MiMeDB 1.0. MiMeDB 2.0’s pathways illustrate not only the origins of most microbial metabolites found in MiMeDB 2.0, but also their subsequent transformations (including those driven by the human host), providing essential mechanistic context. In addition to its collection of over 10 million metabolic pathways, MiMeDB also includes nearly 460 000 metabolite signaling, quorum sensing, and metabolite activation (i.e. operon) pathways.

All pathways in MiMeDB 2.0 were generated using PathWhiz [26] and integrated into PathBank [25] and PathBacK (https://pathback.wishartlab.com/), where they are displayed as interactive pathway maps. Considerable effort was devoted to manually researching and constructing 12 996 primary or “canonical” pathways for bacterial (both archaebacterial and eubacterial) metabolism, which were then systematically propagated across other microbial species. This was achieved by performing BLAST searches against the UniProt database [10], using enzyme or protein sequences from the canonical pathway to identify homologous proteins in additional bacterial species. This “propagation” step has been explained in detail previously [25] and makes use of a number of sequence/nomenclature checks and balances to ensure that all functional annotation and pathway transfers are well justified, modest in scope, and conform to standard requirements. These methods also align closely with the annotation transfer methods used by KEGG and BioCyc [5, 6]. Through this annotation transfer approach, 23.1 million additional, bacterial-specific pathways were computationally inferred, enabling broad coverage of microbial metabolism across nearly all bacterial species represented in MiMeDB. A similar annotation transfer method and pathway propagation process was implemented for all fungal genomes in MiMeDB 2.0. For this fungal pathway propagation process, the canonical pathways for Saccharomyces cerevisiae found in PathBank 2.0 [25] and YMDB [15] were primarily used.

Prokariotic and eukariotic genome viewers

As noted earlier, all bacterial genomes in MiMeDB were annotated using BASys2 [18], which provides accurate and comprehensive gene information for integration into genome maps. These maps were incorporated into MiMeDB’s interactive Prokaryotic Genome Viewer (Fig. 2), allowing users to explore genomic features at multiple levels of detail. Selecting a gene or metabolite opens a MiMeDB GeneCard “pop-up” containing up to 62 annotation fields, including gene/protein name, type, synonyms, genomic coordinates, DNA length, and adjacent genes. For protein-coding features, additional details such as protein length, molecular weight, amino acid sequence, functional assignments, motifs, and database cross-references are provided. Predicted protein secondary structure and 3D structures can also be visualized interactively, and both DNA and protein sequences are downloadable in FASTA format. Each annotation is linked to metadata describing the prediction source and confidence score.

Figure 2.

Figure 2.

MiMeDB 2.0 Genome Viewer. Genome annotations are displayed in a circular view. (A) Interactive Bacterial Genome Viewer with a searchable gene table below, where the “eye” icon links entries to the genome viewer. (B) Eukaryotic Genome Viewer segmented by chromosomes.

To complement the Prokaryotic Genome Viewer, MiMeDB 2.0 also includes a Eukaryotic Genome Viewer based on EUKLID annotations. EUKLID was used to annotate genomes from 84 fungal species, of which 77 had unique complete or chromosome-level assemblies, including 53 from RefSeq [28] and 24 from GenBank [8]. RefSeq genomes contained full metadata and gene feature annotations, whereas some GenBank entries provided only partial annotation. To standardize this data, a Python-based pipeline was developed to extract gene information from NCBI Datasets and convert this information into JSON files compatible with the Eukaryotic Genome Viewer.

EUKLID uses a refactored version of the JavaScript library originally used in BASys2, adapted to accommodate multiple chromosomes and permit visualization of exon/intron boundaries in a circular view. Chromosomes are displayed as contiguous segments around a circle, separated by gaps, enabling clear visualization of their organization, along with their GC content and GC skew. Each chromosome is assigned a distinct color and textual label, allowing users to easily distinguish genomic features across contigs. EUKLID’s unique circular view allows a more compact representation and an easier comparison of eukaryotic genomic data, making it ideal for displaying CIRCOS-like plots [32].

A comprehensive gene table is displayed below each Genome Viewer (both eukaryotic and prokaryotic). Users can search the selected genome by gene name or perform BLAST queries with DNA or protein sequences against the genome. Table entries link directly to the corresponding MiMeDB GeneCard for full annotation details, while an “eye” icon can be used by users to zoom the viewer into the selected gene. This ensures seamless integration between search results, annotations, and genomic context.

Added exposure sources and relations to metabolites

Exposure sources are a key feature of MiMeDB, as they provide critical context for understanding about metabolite origins and how they enter the human body. In MiMeDB 1.0, most exposure data were derived from FooDB (https://foodb.ca), which primarily covers dietary sources. In MiMeDB 2.0, this coverage has been substantially expanded by incorporating data from the U.S. Food and Drug Administration’s (FDA) Substances Added to Food database (www.hfpappexternal.fda.gov/scripts/fdcc/index.cfm?set=FoodSubstances). This update links metabolites to specific categories of FDA-regulated compounds, including food additives and color additives, substances designated as “Generally Recognized as Safe,” and compounds approved for defined uses in foods.

Beyond food exposures, MiMeDB 2.0 now includes additional sources such as drugs, personal care products, household products, water, soil, and other components of the human exposome. Each metabolite has been cross-referenced and matched to its corresponding FDA- or literature-derived source, ensuring more complete coverage of environmental and anthropogenic exposures. To make this information easier to navigate, exposure data are displayed in two tables: one focused on food-related exposures and a second highlighting non-food sources. MiMeDB’s food table now shows specific species (plant, animal, fungi) in a clickable pop-up to make the table easier to read. These tables are accessible in the MetaboCard under the “Exposure Sources” data field, where two tables are displayed. Specific metabolites originating from a given source can be located either by using MiMeDB’s metabolite filter or by browsing the exposure sources table in the browse menu and sorting entries by exposure source. With the rapid advancements in exposome research, such links will continue to be added to the database.

Better spectral data and improved spectral searches

MS and NMR spectral data [along with chromatographic retention time (RT) data] are critical for metabolite identification and validation. One of the intended uses of MiMeDB is to assist researchers in the identification of microbial metabolites or microbially derived metabolites found in stool samples, urine, or even in blood. In MiMeDB 2.0, spectral coverage has been extensively expanded by integrating all experimental and predicted datasets from HMDB 5.0 [13]. For shared metabolites, entries were revised to provide complete and harmonized spectral information.

New, higher-quality MS/MS (tandem mass) spectral predictions were generated for all compounds in MiMeDB that lacked high-quality experimental MS/MS spectra. These predictions were done using FraGGNet v2 [21], which has been shown to be ~50% more accurate than MS/MS spectra predicted via CFM-ID [33] (used previously in MiMeDB 1.0). FraGNNet predictions were generated for multiple collision energies (10, 20, and 40 eV) and for both positive and negative ionization modes. In addition to MS/MS data, EI-MS (electron impact MS) data were predicted using a locally developed version of NEIMS [22] for all MiMeDB compounds with a molecular weight below 900 daltons. This EI-MS prediction program also performs trimethylsilylation and generates all chemically reasonable TMS derivatives for a given starting compound. To assist with compound identification by MS methods, MiMeDB 2.0 also includes RT data generated via RT-Pred [19] for 19 common chromatographic methods and retention index (RI) data generated via RIPred [20]. Both RT-Pred and RIPred are based on specially trained machine learning algorithms, and both have been shown to have the highest accuracy among published programs for RI and RT prediction [22, 34, 35].

MiMeDB 2.0 also includes extensive NMR spectral data, much of it predicted by CASPRE (https://caspre.ca/) for 13C spectra and PROSPRE [36] for 1H spectra. Combinations of the two predictors were used to generate 13C-1H 2D NMR spectra. CASPRE and PROSPRE are based on machine learning (GNN or graphical neural network) algorithms, and both programs have consistently been shown to have the highest accuracy of known chemical shift predictors, especially for compounds dissolved in water (the most common solvent used in metabolomics).

On the experimental side, MiMeDB 2.0 was further enriched with close to 17 000 experimental EI-MS and MS/MS spectra, corresponding to 890 compounds have been added from MassBank [37]. Similarly, another 208 819 predicted 1H NMR spectra, 20 894 predicted 13C NMR spectra, and 27 022 predicted 13C-1H NMR spectra have been added (covering various NMR spectrometer frequencies). Additionally, 3300 experimentally measured IR spectra have been added. A new version of JSpectraView [24] was implemented to permit rapid, interactive, and consistent display of all the MiMeDB 2.0 spectra. Users may access and view compound-specific spectra or chromatographic data from individual MiMeDB 2.0 MetaboCards using the various Browse or Search options available at the top of the MiMeDB 2.0 homepage.

MiMeDB 2.0’s spectral search tools have also been significantly upgraded, enabling much faster (∼5 s in MiMeDB 2.0 compared to ∼10 s in MiMeDB 1.0) and much more accurate similarity, structure, and substructure queries that link directly to MiMeDB 2.0’s metabolite entries. For the MS spectral searches, the MassQL [38] program was integrated into the MiMeDB 2.0’s search function backend. This implementation vastly improved the search speed, moving search times from a logarithmic runtime dependence [O(log)] to a near-constant runtime dependence [O(1)]. Similar improvements were made to the NMR spectral searches. MiMeDB 2.0’s search interface can now handle direct XML input files (such as mzML or nmrML), rather than raw peak lists, improving the usability of its spectral search tool. Together, these updates make MiMeDB 2.0 one of the most comprehensive resources for microbial metabolite spectral data, as well as one of the most powerful metabolite identification tools for microbial metabolomics.

Improved layout, design, and usability

Database search and filter improvements

MiMeDB contains 15.5 gigabytes of data, covering 3725 microbes and 29 295 metabolites, as well as millions of genes and of pathways. This makes it an extraordinarily large database that can be intrinsically difficult to navigate. To address these issues, MiMeDB 2.0 has introduced a number of advanced file/data filtering and refined search tools to streamline access to its many metabolites, microbes, genes, pathways, and reactions (Fig. 3). Metabolites can now be filtered by origin, structure class, host, microbial metabolism, health outcome, bioactivity, biospecimen, and detection status. The addition of “metabolite origin” replaces the older “metabolite type” filter, offering more specific and informative options. Microbes can be filtered by host, body site, biospecimen, taxonomy, pathogenicity, energy production, oxygen preference, Gram stain, cell shape, and arrangement.

Figure 3.

Figure 3.

Metabolites browse page filtered by health outcome and biospecimen. The filtered results can be exported as a CSV file using the “Download Filtered” button.

In addition to the filter function, a general search bar located in the top-right corner of each MiMeDB page allows users to search microbes and their descriptions using simple text queries. A more advanced search option is available under the “Search” dropdown menu, where the “Microbe Search” tool enables queries across all of the above categories, as well as searching microbes by genus and species. Reaction searches now include a filter for microbial species, enabling species-specific exploration. In addition to these enhancements, all filters in MiMeDB 2.0 have been reorganized for greater clarity and ease of use. Filtered datasets can now be directly exported in CSV format, allowing users to download and analyze targeted subsets of data. An example of a filtered dataset is shown in Fig. 3. Query performance has also been significantly optimized, with typical response times now reduced to ~1–5 s (previously ∼2–20 s on MiMeDB 1.0) for simple queries, and ~2–10 s (previously 5–60+ s) for complex queries. For better accessibility of microbe/metabolite-specific data, a download button has also been added to the microbe page that allows users to download a CSV file of all related metabolites of that microbe.

Network viewer

MiMeDB’s Network Viewer is an interactive tool for visualizing relationships among metabolites, microbes, associated health effects, biospecimens, bioactivities, and exposure sources (Fig. 4). The Network Viewer employs Data-Driven Documents or D3.js version 7 [39], which is an open-source JavaScript library for creating dynamic and interactive data visualizations in web browsers. It allows developers to bind data to the Document Object Model and then manipulate those elements based on the data. In particular, D3.js can be used to build network diagrams by combining force-directed graph layouts with dynamic SVG rendering. It positions nodes and links using the “d3.forceSimulation” function or subroutine, updates their coordinates, and binds data-driven attributes like color, size, and labels. This enables fully interactive, animated, and scalable network visualizations from raw graph data. In this way users can use a simple query to easily generate dynamic, web-based graphical network diagrams and interactively explore connections by dragging, expanding, and highlighting nodes.

Figure 4.

Figure 4.

Network Viewer in MiMeDB 2.0. (A) Selecting a metabolite updates the microbe dropdowns to reflect its relationships. (B) Network Viewer view with Indoxyl Sulfate highlighted, spreading out the nodes to improve readability. (C) Network Viewer with ‘Health outcome’ selected in the legend, highlighting only health outcomes.

To initiate a Network Viewer query, users can select a metabolite from an autocomplete list, after which only relevant microbes are displayed. Up to five different microbes can be chosen, and users can specify the degree of connectivity to control the level of network complexity. Lower values (between three and five) produce readable views, while higher values tend to generate somewhat more dense networks, which are ideal for interactive exploration.

The resulting MiMeDB Network Viewer visualization displays different colored nodes for metabolites, microbes, associated health effects, biospecimens, bioactivities, and/or exposure sources along with their associated links. Interactive features include hover-based highlighting of immediate connections, drag-and-drop repositioning, and category-based highlighting through the legend. Clicking a node links directly to the corresponding database entry. Collectively, these improvements enhance the viewer’s readability, interactivity, and integration, making the MiMeDB Network Viewer a more effective tool for navigating MiMeDB 2.0’s complex data.

Database implementation, curation, and fairness

MiMeDB 2.0 was developed using the Ruby on Rails framework (https://rubyonrails.org, version 2.5.1) with a SQL backend, following a Model–View–Controller design to ensure modularity and efficient data handling. The database is deployed on a dedicated Ubuntu 22.04 server within OVHcloud (www.ovhcloud.com/en-ca/), with containerization to safeguard performance across shared resources. As highlighted earlier, all MiMeDB entries were manually curated and validated by a trained team of biologists and biochemists. Data entry was conducted through a centralized, password-controlled system, with all changes logged, time-stamped, and verified to maintain quality and consistency. Additional details about the MiMeDB 2.0’s curation process and quality management program were provided in the original MiMeDB manuscript [40].

MiMeDB 2.0 fully adheres to FAIR principles: Findable, Accessible, Interoperable, and Reusable [41]. Each microbe, metabolite, and reaction is assigned a unique and permanent MMDB identifier. Data are accessible via user-friendly web pages and bulk downloads, offered in standard formats such as CSV and XML. Careful curation, extensive documentation, and standardized ontologies ensure that the database remains interoperable and reusable for diverse applications. Additional details regarding MiMeDB’s adherence to FAIRness are available under the “About” menu.

Conclusion and future directions

MiMeDB 2.0 substantially expands the size and scope of the MiMeDB, adding thousands of new metabolites, microbes, reactions, and spectra as well as millions of new genes, proteins, annotations, and pathways to its holdings. In addition to these enormous data expansions, significant effort went into improving the usability of the database. These include much improved viewing tools such as the Prokaryotic Genome Viewer, the Eukaryotic Genome Viewer, the MiMeDB Network Viewer, and JSpectraView. They also include significantly improved data filters, more informative data tagging, richer annotations or descriptions of microbes, higher-quality genome/proteome annotations, faster spectral search functions, improved design and layout as well as faster and more stable performance. Perhaps the most important paradigm-shifting development with MiMeDB 2.0 is the fact that it is now the first complete, end-to-end, multi-omics database. MiMeDB 2.0 brings full visualization, searching, and browsing capabilities for genomes, proteomes, metabolomes, microbiomes, pathways, and network diagrams. No other online database that we are aware of offers this depth of multi-omic support. Substantial qualitative and quantitative improvements to MiMeDB 2.0’s front-end and back-end have given it those end-to-end capabilities. Together, we believe these enhancements strengthen MiMeDB’s role as a central resource for exploring the biochemical interface between humans and their microbiota.

As microbiome research continues to advance at a rapid pace, new microbes, new genomes, new metabolites, and new health or dietary associations are being identified almost weekly, underscoring the need for regular updates and frequent curation. The size and frequency of these publications now exceed our capacity to manually keep MiMeDB as current as we would like. Going forward, our team will increasingly have to rely on customized ML-aided methods, such as custom LLMs and other natural language processing tools, as well as other customized semi-automated annotation tools to handle the flood of published (and electronic) data. We have begun to develop and refine some of these tools in this release of MiMeDB and are encouraged by their general success. We will also rely on user feedback on data or any issues, which can be given through the “Contact Us” page, which can be accessed through the navigation bar at the top of each MiMeDB page. As with any database project, we strive to ensure the sustainability, currency, and reliability of MiMeDB. Our hope is that the ideas and methods described here will allow MiMeDB to remain comprehensive, reliable, and FAIR-compliant and that it will continue to be improved by integrating new findings and adding better, more powerful user-facing tools. Our hope is that with continued expansion of the database across genomic, proteomic, metabolomic, and exposomic dimensions, MiMeDB will serve as a key platform for both understanding and discovering the roles of microbes and microbial metabolites in human health, diet, and disease.

Acknowledgements

We thank the UCSD Dorrestein Lab team for providing early, unpublished access to the Collaborative Microbial Metabolite Center Knowledgebase.

Author contributions: Ray Kruger (Data curation [lead], Resources [lead], Software [equal], Writing—original draft [lead]), Eponine Oler (Data curation [supporting], Resources [supporting], Software [supporting], Validation [lead], Visualization [supporting]), Sukanta Saha (Data curation [supporting], Resources [supporting], Software [supporting]), Jenna Poelzer (Data curation [supporting], Software [supporting]), Scott Han (Software [supporting], Visualization [supporting]), Brittany Green (Data curation [supporting], Visualization [supporting]), Fahrin Bushra (Data curation [supporting]), Megane Kyes (Data curation [supporting], Resources [supporting], Visualization [supporting]), Fehintola Disu (Resources [supporting], Visualization [supporting]), Julia Wakoli (Resources [supporting], Visualization [supporting]), Robyn Woudstra (Data curation [supporting], Resources [supporting], Software [supporting]), Elykah Tejol (Data curation [supporting], Resources [supporting], Validation [supporting]), Fei Wang (Data curation [supporting], Resources [supporting]), Brian Lee (Data curation [supporting], Resources [supporting]), Tanvir Sajed (Software [supporting]), Siyang Tian (Data curation [supporting], Software [supporting]), Claudia Torres-Calzada (Writing—original draft [supporting], Writing—review & editing [supporting]), Mark Berjanskii (Supervision [supporting], Validation [supporting]), Scott MacKay (Supervision [supporting], Validation [supporting], Visualization [supporting]), Naama Karu (Data curation [supporting], Supervision [supporting]), and David S. Wishart (Conceptualization [lead], Data curation [supporting], Formal analysis [supporting], Funding acquisition [lead], Methodology [lead], Project administration [lead], Resources [lead], Software [supporting], Supervision [lead], Validation [lead], Visualization [supporting], Writing—original draft [equal], Writing—review & editing [lead]).

Contributor Information

Ray Kruger, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Eponine Oler, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Sukanta Saha, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada; Department of Computing Science, University of Alberta, Edmonton, AB T6G 2E8, Canada.

Jenna Poelzer, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Scott Han, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Mark Peter Punsalan, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Brittany Green, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Fahrin Bushra, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Megane Kyes, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Fehintola Disu, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Julia Wakoli, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Robyn Woudstra, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Elykah Tejol, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Omolola Fakutan, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Fei Wang, Department of Computing Science, University of Alberta, Edmonton, AB T6G 2E8, Canada.

Brian L Lee, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Tanvir Sajed, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Siyang Tian, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Claudia Torres-Calzada, Department of Medical Microbiology and Immunology, University of Alberta, Edmonton, AB T6G 2E1, Canada.

Mark Berjanskii, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Scott MacKay, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada.

Naama Karu, Tasmanian Independent Metabolomics and Analytical Chemistry Solutions (TIMACS), Hobart, Tasmania 7008, Australia.

Rima Kadurrah-Daouk, Department of Medicine, Duke University School of Medicine, Durham, NC 27710, United States.

David S Wishart, Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada; Department of Computing Science, University of Alberta, Edmonton, AB T6G 2E8, Canada; Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, AB T6G 2B7, Canada; Department of Pharmacy and Pharmaceutical Sciences, University of Alberta, Edmonton, AB T6G 2H7, Canada.

Conflict of interest

None declared.

Funding

This work was supported by the National institute of Health (NIH) National Institute of Aging (NIA); U19 [AG063744/AG/NIA NIH HHS/United States]; Canada Foundation for Innovation (CFI MSIF 42495); Canada Research Chairs (CRC TIER1 100628); Genome Alberta, a division of Genome Canada; The Natural Sciences and Engineering Research Council of Canada (NSERC) grant ALLRP 597259-24; the Weston Family Foundation (UBC F23-05862 WFF); and the Social Sciences and Humanities Research Council of Canada (SSHRC) grant 2007-2024-1003. Funding to pay the Open Access publication charges for this article was provided by Genome Alberta.

Data availability

All data from MiMeDB 2.0 are freely accessible through the MiMeDB 2.0 website or the dedicated MiMeDB 2.0 download page. The platform is open and free to use, with data downloads compatible with all modern web browsers. Download files are available in CSV and XML formats. Additionally, metabolites can be accessed in multiple formats, including SMILES, SDF, MOL, PDB, InChI, and InChIKey. Gene sequence data are stored and available in FASTA format. Images are available in PNG format, and all textual data and metadata are written in English. The data in MiMeDB are released under the Creative Commons (CC) 4.0 License Suite according to the Attribution BY and Non-Commercial NC licensing conditions.

References

  • 1. Bishai JD, Palm NW. Small molecule metabolites at the host-microbiota interface. J Immunol. 2021;207:1725–33. 10.4049/jimmunol.2100528. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Singh V, Lee G, Son Het al. Butyrate producers, “The Sentinel of Gut”: their intestinal significance with and beyond butyrate, and prospective use as microbial therapeutics. Front Microbiol. 2023;13:1103836. 10.3389/fmicb.2022.1103836. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Ye X, Li H, Anjum Ket al. Dual role of indoles derived from intestinal microbiota on human health. Front Immunol. 2022;13:903526. 10.3389/fimmu.2022.903526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Yeh YC, Huang MF, Liang SSet al. Indoxyl sulfate, not p-cresyl sulfate, is associated with cognitive impairment in early-stage chronic kidney disease. Neurotoxicology. 2016;53:148–52. 10.1016/j.neuro.2016.01.006. [DOI] [PubMed] [Google Scholar]
  • 5. Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000;28:27–30. 10.1093/nar/28.1.27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Karp PD, Billington R, Caspi Ret al. The BioCyc collection of microbial genomes and metabolic pathways. Brief Bioinform. 2019;20:1085–93. 10.1093/bib/bbx085. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Noronha A, Modamio J, Jarosz Yet al. The Virtual Metabolic Human database: integrating human and gut microbiome metabolism with nutrition and disease. Nucleic Acids Res. 2019;47:D614–24. 10.1093/nar/gky992. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Benson DA, Karsch-Mizrachi I, Lipman DJet al. GenBank. Nucleic Acids Res. 2008;36:D25–30. 10.1093/nar/gkm929. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Almeida A, Nayfach S, Boland M.et al. A unified catalog of 204,938 reference genomes from the human gut microbiome. Nat Biotechnol. 2021; 39:105–14. 10.1038/s41587-020-0603-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Apweiler R, Bairoch A, Wu CHet al. UniProt: the Universal Protein knowledgebase. Nucleic Acids Res. 2004;32:D115–9. 10.1093/nar/gkh131. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Cruz J, Liu Y, Liang Yet al. BacMap: an up-to-date electronic atlas of annotated bacterial genomes. Nucleic Acids Res. 2012;40:D599–604. 10.1093/nar/gkr1105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Dyer SC, Austine-Orimoloye O, Azov AGet al. Ensembl 2025. Nucleic Acids Res. 2025;53:D948–57. 10.1093/nar/gkae1071. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Wishart DS, Guo A, Oler Eet al. HMDB 5.0: the Human Metabolome Database for 2022. Nucleic Acids Res. 2022;50:D622–31. 10.1093/nar/gkab1062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Sajed T, Marcu A, Ramirez Met al. ECMDB 2.0: a richer resource for understanding the biochemistry of E. coli. Nucleic Acids Res. 2016;44:D495–501. 10.1093/nar/gkv1060. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Jewison T, Knox C, Neveu Vet al. YMDB: the Yeast Metabolome Database. Nucleic Acids Res. 2012;40:D815–20. 10.1093/nar/gkr916. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Jackson H, Oler E, Torres-Calzada Cet al. MarkerDB 2.0: a comprehensive molecular biomarker database for 2025. Nucleic Acids Res. 2025;53:D1415–26. 10.1093/nar/gkae1056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Schober I, Koblitz J, Sardà Carbasse Jet al. BacDive in 2025: the core database for prokaryotic strain data. Nucleic Acids Res. 2025;53:D748–56. 10.1093/nar/gkae959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Poelzer J, Han S, Saha Set al. BASys2: a next-generation bacterial genome annotation system. Nucleic Acids Res. 2025;53:W57–67. 10.1093/nar/gkaf360. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Zakir M, LeVatte MA, Wishart DS. RT-Pred: a web server for accurate, customized liquid chromatography retention time prediction of chemicals. J Chromatogr A. 2025;1747:465816. 10.1016/j.chroma.2025.465816. [DOI] [PubMed] [Google Scholar]
  • 20. Anjum A, Liigand J, Milford Ret al. Accurate prediction of isothermal gas chromatographic Kováts retention indices. J Chromatogr A. 2023;1705:464176. 10.1016/j.chroma.2023.464176. [DOI] [PubMed] [Google Scholar]
  • 21. Young A, Wang F, Wishart Det al. FraGNNet: a deep probabilistic model for mass spectrum prediction. arXiv, https://arxiv.org/abs/2404.02360, 2 April 2024, preprint: not peer reviewed.
  • 22. Wakoli J, Anjum A, Sajed Tet al. GCMS-ID: a webserver for identifying compounds from gas chromatography mass spectrometry experiments. Nucleic Acids Res. 2024;52:W381–9. 10.1093/nar/gkae425. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Sajed T, Sayeeda Z, Lee BLet al. Accurate prediction of 1H NMR chemical shifts of small molecules using machine learning. Metabolites. 2024;14:290. 10.3390/metabo14050290. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Schober D, Jacob D, Wilson Met al. nmrML: a community supported open data standard for the description, storage, and exchange of NMR data. Anal Chem. 2018;90:649–56. 10.1021/acs.analchem.7b02795. [DOI] [PubMed] [Google Scholar]
  • 25. Wishart DS, Kruger R, Sivakumaran Aet al. PathBank 2.0-the pathway database for model organism metabolomics. Nucleic Acids Res. 2024;52:D654–62. 10.1093/nar/gkad1041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Pon A, Jewison T, Su Yet al. Pathways with PathWhiz. Nucleic Acids Res. 2015;43:W552–9. 10.1093/nar/gkv399. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Bansal P, Morgat A, Axelsen KBet al. Rhea, the reaction knowledgebase in 2022. Nucleic Acids Res. 2022;50:D693–700. 10.1093/nar/gkab1016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Goldfarb T, Kodali VK, Pujar Set al. NCBI RefSeq: reference sequence standards through 25 years of curation and annotation. Nucleic Acids Res. 2025;53:D243–57. 10.1093/nar/gkae1038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Galperin MY, Karamycheva Set al. COG database update 2024. Nucleic Acids Res. 2025;53:D356–63. 10.1093/nar/gkae983. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Vu TTD, Jung J. Protein function prediction with gene ontology: from traditional to deep learning models. PeerJ. 2021;9:e12019. 10.7717/peerj.12019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Yang Z, Zeng X, Zhao Yet al. AlphaFold2 and its applications in the fields of biology and medicine. Sig Transduct Target Ther. 2023;8:115. 10.1038/s41392-023-01381-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Naquin D, d’Aubenton-Carafa Y, Thermes Cet al. CIRCUS: a package for Circos display of structural genome variations from paired-end and mate-pair sequencing data. BMC Bioinformatics. 2014;15:15–198. 10.1186/1471-2105-15-198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Wang F, Allen D, Tian Set al. CFM-ID 4.0 - a web server for accurate MS-based metabolite identification. Nucleic Acids Res. 2022;50:W165–74. 10.1093/nar/gkac383. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Wang QY, Zhu YL, Li XH.et al. Construction of a machine learning ensemble prediction model for gas chromatographic retention index on stationary phases with different polarities. Chin J Chromatogr. 2025;43:355–62. 10.3724/SP.J.1123.2024.07014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Knox C, Wilson M, Klinger CMet al. DrugBank 6.0: the DrugBank knowledgebase for 2024. Nucleic Acids Res. 2024;52:D1265–75. 10.1093/nar/gkad976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Sajed T, Sayeeda Z, Lee BLet al. Accurate prediction of 1H NMR chemical shifts of small molecules using machine learning. Metabolites. 2024;14:290. 10.3390/metabo14050290. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Horai H, Arita M, Kanaya Set al. MassBank: a public repository for sharing mass spectral data for life sciences. J Mass Spectrom. 2010;45:703–14. 10.1002/jms.1777. [DOI] [PubMed] [Google Scholar]
  • 38. Damiani T, Jarmusch AK, Aron ATet al. A universal language for finding mass spectrometry data patterns. Nat Methods. 2025;22:1247–54. 10.1038/s41592-025-02660-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Afonso MQL, NJ dFJ, Miranda TGet al. A d3.js based JavaScript library for drawing and annotating voltage-gated sodium channels membrane diagrams. Front Bioinform. 2022;2:774417. 10.3389/fbinf.2022.774417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Wishart DS, Oler E, Peters Het al. MiMeDB: the Human Microbial Metabolome Database. Nucleic Acids Res. 2023;51:D611–20. 10.1093/nar/gkac86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Wilkinson MD, Dumontier M, Aalbersberg IJet al. The FAIR guiding principles for scientific data management and stewardship. Sci Data. 2016;3:160018. 10.1038/sdata.2016.18. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

All data from MiMeDB 2.0 are freely accessible through the MiMeDB 2.0 website or the dedicated MiMeDB 2.0 download page. The platform is open and free to use, with data downloads compatible with all modern web browsers. Download files are available in CSV and XML formats. Additionally, metabolites can be accessed in multiple formats, including SMILES, SDF, MOL, PDB, InChI, and InChIKey. Gene sequence data are stored and available in FASTA format. Images are available in PNG format, and all textual data and metadata are written in English. The data in MiMeDB are released under the Creative Commons (CC) 4.0 License Suite according to the Attribution BY and Non-Commercial NC licensing conditions.


Articles from Nucleic Acids Research are provided here courtesy of Oxford University Press

RESOURCES