CARD 2023: expanded curation, support for machine learning, and resistome prediction at the Comprehensive Antibiotic Resistance Database.
Alcock BP, Huynh W, Chalil R, Smith KW, Raphenya AR, Wlodarski MA, Edalatmand A, Petkau A, Syed SA, Tsang KK, Baker SJC, Dave M, McCarthy MC, Mukiri KM, Nasir JA, Golbon B, Imtiaz H, Jiang X, Kaur K, Kwong M, Liang ZC, Niu KC, Shan P, Yang JYJ, Gray KL, Hoad GR, Jia B, Bhando T, Carfrae LA, Farha MA, French S, Gordzevich R, Rachwalski K, Tu MM, Bordeleau E, Dooley D, Griffiths E, Zubyk HL, Brown ED, Maguire F, Beiko RG, Hsiao WWL, Brinkman FSL, Van Domselaar G, McArthur AG. · 2023
Read CARD 2023: expanded curation, support for machine learning, and resistome prediction at the Comprehensive Antibiotic Resistance Database. by Alcock BP, Huynh W, Chalil R, Smith KW, Raphenya AR, Wlodarski MA, Edalatmand A, Petkau A, Syed SA, Tsang KK, Baker SJC, Dave M, McCarthy MC, Mukiri KM, Nasir JA, Golbon B, Imtiaz H, Jiang X, Kaur K, Kwong M, Liang ZC, Niu KC, Shan P, Yang JYJ, Gray KL, Hoad GR, Jia B, Bhando T, Carfrae LA, Farha MA, French S, Gordzevich R, Rachwalski K, Tu MM, Bordeleau E, Dooley D, Griffiths E, Zubyk HL, Brown ED, Maguire F, Beiko RG, Hsiao WWL, Brinkman FSL, Van Domselaar G, McArthur AG. (2023) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc9825576.txt.
Introduction
Alcock and co-authors report the 2023 state of CARD, a curated database of antibiotic resistance genes. It pairs a shared scheme for naming resistance traits with gene records and the mutations that confer them. Software built on the records reads new genomes and names the resistance they carry.
The build holds 6,627 terms in its naming scheme and 5,010 reference gene sequences. It holds 1,933 curated mutations and 5,057 detection models for the partner search tool. Curation widened on several fronts since 2020, with deeper coverage of a major drug breaking enzyme family. Records for tuberculosis drug resistance grew, and disinfectant resistance joined the scope of the database. Resistance blocking agents were curated in a systematic pass as well. Predicted resistance profiles now span 377 pathogen species in the system. Each gene also gained a short standard name to support machine learning work. A live collection gathers resistance data sent in by the wider community.
Catalog record
| CREATOR | Alcock BP, Huynh W, Chalil R, Smith KW, Raphenya AR, Wlodarski MA, Edalatmand A, Petkau A, Syed SA, Tsang KK, Baker SJC, Dave M, McCarthy MC, Mukiri KM, Nasir JA, Golbon B, Imtiaz H, Jiang X, Kaur K, Kwong M, Liang ZC, Niu KC, Shan P, Yang JYJ, Gray KL, Hoad GR, Jia B, Bhando T, Carfrae LA, Farha MA, French S, Gordzevich R, Rachwalski K, Tu MM, Bordeleau E, Dooley D, Griffiths E, Zubyk HL, Brown ED, Maguire F, Beiko RG, Hsiao WWL, Brinkman FSL, Van Domselaar G, McArthur AG. |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
CARD 2023: expanded curation, support for machine learning, and resistome prediction at the Comprehensive Antibiotic Resistance Database By Brian P Alcock, William Huynh, Romeo Chalil, Keaton W Smith, Amogelang R Raphenya, Mateusz A Wlodarski, Arman Edalatmand, Aaron Petkau, Sohaib A Syed, Kara K Tsang, Sheridan J C Baker, Mugdha Dave Published in: Nucleic Acids Research DOI: 10.1093/nar/gkac920 ABSTRACT Abstract The Comprehensive Antibiotic Resistance Database (CARD; card.mcmaster.ca) combines the Antibiotic Resistance Ontology (ARO) with curated AMR gene (ARG) sequences and resistance-conferring mutations to provide an informatics framework for annotation and interpretation of resistomes. As of version 3.2.4, CARD encompasses 6627 ontology terms, 5010 reference sequences, 1933 mutations, 3004 publications, and 5057 AMR detection models that can be used by the accompanying Resistance Gene Identifier (RGI) software to annotate genomic or metagenomic sequences. Focused curation enhancements since 2020 include expanded β-lactamase curation, incorporation of likelihood-based AMR mutations for Mycobacterium tuberculosis, addition of disinfectants and antiseptics plus their associated ARGs, and systematic curation of resistance-modifying agents. This expanded curation includes 180 new AMR gene families, 15 new drug classes, 1 new resistance mechanism, and two new ontological relationships: evolutionary_variant_of and is_small_molecule_inhibitor. In silico prediction of resistomes and prevalence statistics of ARGs has been expanded to 377 pathogens, 21,079 chromosomes, 2,662 genomic islands, 41,828 plasmids and 155,606 whole-genome shotgun assemblies, resulting in collation of 322,710 unique ARG allele sequences. New features include the CARD:Live collection of community submitted isolate resistome data and the introduction of standardized 15 character CARD Short Names for ARGs to support machine learning efforts. INTRODUCTION In the decade since the initial publication of the Comprehensive Antibiotic Resistance Database (CARD) (1), the public health threat of antimicrobial resistance (AMR) has grown significantly, prompting a collective global initiative to combat the spread of AMR through collaborative research, molecular and phenotypic surveillance, and antimicrobial stewardship (2–5). However, the complexity and global burden of the AMR crisis (6) continues to expand with the continuous discovery of novel antibiotic resistance genes (ARGs), including those using novel mechanisms such as MCR-mediated colistin resistance via change in the charge of the cell membrane (7) or ejection of rifamycin from its target RNA polymerase by helicase-like protein HelR (8), as well as contributing factors from environmental, veterinary, and food-related sources (9,10). Furthermore, emerging evidence details a complex relationship between AMR and the COVID-19 pandemic, in part due to prolonged hospitalizations, relaxed antibiotic stewardship in overwhelmed hospital settings, and increased household and topical biocide use (11,12). As previously described (13), CARD is an ontology-centric knowledgebase on the molecular determinants of antibiotic resistance. By integrating a detailed controlled vocabulary, the Antibiotic Resistance Ontology (ARO), with ARG molecular sequences and mutation data, CARD is able to provide both reference set and software tools for guiding AMR research, particularly for ARG annotation and discovery from genomic and metagenomic data. One such analytical framework is the Resistance Gene Identifier (RGI), CARD’s algorithm for computational AMR genotype and phenotype prediction from genomic data using the bioinformatics detection models curated in CARD. To ensure accuracy, CARD is manually curated by a team of biocurators and researchers, following a core curation paradigm. Additionally, CARD works towards harmonization with other AMR databases, specifically including those provided by the National Center for Biotechnology Information (NCBI), such as the Pathogen Detection Reference Gene Catalog (https://www.ncbi.nlm.nih.gov/pathogens/). Here we provide an update on the state of CARD and the ARO, with a review of the available data and updates since our 2020 publication. We discuss in detail (i) general and specific improvements to the ARO and curation of AMR in CARD, (ii) improved integration and harmonization with existing resources for specific AMR gene families (such as β-lactamase resistance genes) and pathogens (such as Mycobacterium tuberculosis), (iii) curation of resistance modifying agents, (iv) gene name standardization in support of machine learning efforts, (v) improvements to CARD’s Model Ontology in support of bioinformatic annotation of genomes or metagenomes, (vi) description of CARD’s curation paradigms, protocols and quality assurance algorithms and (vii) expansion of the previously introduced CARD Resistomes & Variants (CARD-R) resource. EXPANSION OF CARD CURATION Current state of CARD and the ARO CARD’s primary curation paradigm continues as previously described: ‘to be included in CARD an AMR determinant must be described in a peer-reviewed scientific publication, with its DNA sequence available in GenBank, including clear experimental evidence of elevated minimum inhibitory concentration (MIC) over controls. AMR genes predicted by in silico methods, but not experimentally characterized, are not included in CARD’s primary curation’ (13). Data in CARD is organized using four core ontologies: the ARO describing all aspects of AMR, the CARD Model Ontology (MO) defining bioinformatics parameters for annotating ARGs in genomic data, the CARD Relations Ontology (RO) for association of terms within ontologies, and NCBITaxon for tracking the source organisms and taxonomic distribution of ARG sequences. The ARO is the primary ontology for organization and interpretation of data within CARD and it contains seven major branches: antibiotic molecule (ARO:1000003), mechanism of antibiotic resistance (ARO:1000002), determinant of antibiotic resistance (ARO:3000000), resistance-modifying agents (ARO:0000076), antibiotic target (ARO:3000708), antibiotic biosynthesis (ARO:3000082), and component of AMR genotypic or phenotypic terminology (ARO:3000045). Each curated ARG must include an ontological connection to the antibiotic molecule, mechanism of antibiotic resistance, and determinant of antibiotic resistance branches of the ARO, as outlined below. Notably, since our last publication the ARO terminology branch has been updated to include terms for phenotypic testing standards for the National Antimicrobial Resistance Monitoring System (NARMS) and National Committee for Clinical Laboratory Standards (NCCLS) to support clinical data harmonization efforts. Connections among terms within the ARO are supported by a set of relation terms from CARD’s Relations Ontology, some of which are generic (e.g. is_a or part_of) while others reflect key aspects of antimicrobial resistance (e.g. confers_resistance_to_antibiotic) (Table 1). With use of these relations, terms within the ARO form a knowledge network for AMR and combined with curation rules support computational approaches for annotation of genome sequences, interpretation of these annotations, and development of powerful curation quality control algorithms, as outlined below. [Table: Ontological relationships in use by CARD within the Antibiotic Resistance Ontology (ARO). New terms since the CARD 2020 publication are in bold] In 2020, CARD v3.0.3 included 4336 terms from the ARO, up from 3567 in our 2017 publication (14), and covering 2923 AMR determinants encoded within CARD detection models (13). As of August 2022, CARD v3.2.4 has expanded to 6627 ARO terms and now includes 5057 AMR detection models covering 5010 reference sequences for resistance determinants (including 1933 curated resistance-associated variant mutations). While the ARO provides a detailed framework for describing antibiotics and AMR, CARD also applies ARO Classification Tags where curators manually ‘tag’ specific ARO terms as particularly informative for interpretation (13): primary tags AMR Gene Family, Drug Class, Resistance Mechanism, Antibiotic and secondary tags Efflux Component, Efflux Regulator, Adjuvant, Antibiotic + Adjuvant (Table 2). CARD’s secondary curation paradigm requires that ‘every curated AMR determinant must have an ontological path including each of the four primary ARO classification tags, i.e. the [AMR Gene Family] to which that determinant belongs, the [Resistance Mechanism], the [Drug Class] to which resistance is conferred, and the specific [Antibiotic] with a demonstrably elevated MIC’ (13). Overall, the curated sequence data in CARD v3.2.4 spans