COG database update: focus on microbial diversity, model organisms, and widespread pathogens

Abstract

The Clusters of Orthologous Genes (COG) database, also referred to as the Clusters of Orthologous Groups of proteins, was created in 1997 and went through several rounds of updates, most recently, in 2014. The current update, available at https://www.ncbi.nlm.nih.gov/research/COG, substantially expands the scope of the database to include complete genomes of 1187 bacteria and 122 archaea, typically, with a single genome per genus. In addition, the current version of the COGs includes the following new features: (i) the recently deprecated NCBI’s gene index (gi) numbers for the encoded proteins are replaced with stable RefSeq or GenBank\ENA\DDBJ coding sequence (CDS) accession numbers; (ii) COG annotations are updated for >200 newly characterized protein families with corresponding references and PDB links, where available; (iii) lists of COGs grouped by pathways and functional systems are added; (iv) 266 new COGs for proteins involved in CRISPR-Cas immunity, sporulation in Firmicutes and photosynthesis in cyanobacteria are included; and (v) the database is made available as a web page, in addition to FTP. The current release includes 4877 COGs. Future plans include further expansion of the COG collection by adding archaeal COGs (arCOGs), splitting the COGs containing multiple paralogs, and continued refinement of COG annotations.

INTRODUCTION

For the past 24 years, the Clusters of Orthologous Genes (COGs) database, also known as Clusters of Orthologous Groups of proteins, has been a popular tool for functional and comparative genomics of bacteria and archaea (1–4). Its relatively small collection of fewer than 5000 clusters of orthologous proteins (COGs) consists of the products of the most widespread bacterial and archaeal genes. These include, for example, ribosomal proteins, universal translation factors, aminoacyl-tRNA synthetases, subunits of the RNA polymerase, F-type and A/V-type ATP synthases, as well as many key metabolic and signaling enzymes.

In all these cases, identification of orthologs by cross-genome comparisons, using the COG-making algorithm to create COGs and the COGMaker method to add COG members from new genomes (5–7) allowed unequivocal functional assignments for proteins from diverse genomes, many of which have never been studied experimentally. These assignments reveal the advantages of functionally annotating a protein family as a whole rather than its individual members. The COG annotations including COG names were deliberately chosen to reflect the functional diversity among the members of the respective COG. Those COGs that included experimentally characterized members with substantially different functions were given composite names to reflect this functional diversity (2–4).

The 2014 release of the COGs provided a major update of the COG names and expanded the coverage to 711 genomes representing all bacterial and archaeal genera that had at least one member with a completely sequenced genome by the end of 2013 (3). Conversely, the three eukaryotic genomes (those of the baker’s yeast Saccharomyces cerevisiae, fission yeast Schizosaccharomyces pombe and the microsporidian Encephalitozoon cuniculi) that had been included in the earlier versions of the COGs were discarded because any attempt to incorporate the vastly increased genomic diversity of eukaryotes was deemed to be outside the scope of the project.

In the new COG release described here, we substantially expand the genome collection covered by the COGs, update the annotation of many COGs, and start adding new COGs, at this time, only a limited number of clusters of orthologs that are linked to certain functionalities and have been described in detail in our recent publications. We also restored the lists of the key pathways and functional groups covered by the COGs, a functionality that was discarded in the previous version. Since the time of the latest COG release, the NCBI has shifted from gene index (gi) numbers for the encoded proteins to GenBank accessions and non-unique RefSeq identifiers, which resulted in a large number of broken links. Accordingly, an important technical task for this COG update was replacing the gi numbers with stable RefSeq or GenBank\ENA\DDBJ coding sequence (CDS) accession numbers. We expect that these amendments substantially increase the utility of the COGs for prokaryote genome analysis and annotation.

KEY CHANGES IN THE 2020 COG UPDATE

Expanded genome coverage

In the current update, the genome coverage of the COG database has been expanded to include representatives of all bacterial and archaeal genera that had complete genome sequences released by 1 April 2019, and a selection of genomes released after that date. Given the rapid progress in microbial genome sequencing, covering all sequenced genomes was clearly untenable: even excluding thousands of unfinished (draft) genomes left >16 000 (as of the start of 2019) complete genomes, as specified in the NCBI descriptions of the respective sequence assemblies. Therefore, selection of the organisms for inclusion in the COGs presented some challenges. To provide reasonable coverage of microbial diversity while keeping the number of organisms manageable, we adopted the following approach.

First, we kept most of the 711 organisms from the previous COG release, removing only some of the duplicate members of the same bacterial genus (Bradyrhizobium, Frankia, Granulicella, Rickettsia, Spiribacter) and keeping other bacterial (Bacillus, Clostridium, Escherichia, Mycobacterium, Mycoplasma, Nostoc, Streptococcus) and archaeal (Pyrococcus, Thermoplasma) genera with two or more members. In the case of the two former Sulfolobus species, one of these, Sulfolobus solfataricus, has been reassigned to the new genus Saccharolobus (8). Name changes have been also registered for 42 other organisms from the 2014 COG list (Supplementary Table S1), including some widely studied model organisms and widespread pathogens. These included reclassification of the Lyme disease-causing Borrelia burgdorferi and some other Borrelia spp. to the new genus Borreliella (9), and of the former Clostridium (Peptoclostridium) difficile to Clostridioides difficile (10; Table 1).

Table 1.

Changes in the names of widely used model organisms from the 2014 COGs

Organism name in the 2014 release	Updated organism name
Bacillus halodurans C-125	Alkalihalobacillus halodurans C-125
Borrelia burgdorferi B31	Borreliella burgdorferi B31
Burkholderia xenovorans LB400	Paraburkholderia xenovorans LB400
Caulobacter crescentus CB15	Caulobacter vibrioides CB15
Chlamydophila pneumoniae CWL029	Chlamydia pneumoniae CWL029
Chlorobium tepidum TLS	Chlorobaculum tepidum TLS
Clostridium difficile 630	Clostridioides difficile 630
Klebsiella pneumoniae 342	Klebsiella variicola 342
Mesorhizobium loti MAFF 303099	Mesorhizobium japonicum MAFF 303099
Methanosaeta concilii GP6	Methanothrix soehngenii GP6
Planctomyces brasiliensis DSM 5305	Rubinisphaera brasiliensis DSM 5305
Sulfolobus solfataricus P2	Saccharolobus solfataricus P2

Second, we added representatives of newly sequenced bacterial and archaeal genera, including those with the Candidatus status. This amendment resulted in the addition of 507 bacterial and 36 archaeal genera. When there were several newly sequenced genomes from the same genus, the choice was typically made in favor of either (i) an organism included in the NCBI representative genome collection (https://www.ncbi.nlm.nih.gov/refseq/about/prokaryotes/#representative_genomes), (ii) the better-studied organism (judging by the number of PubMed references) or (iii) the organism with the largest genome (and, hence, the largest number of encoded proteins). The increase in the number and diversity of complete prokaryote genomes occurred, primarily, via expansion of the already well-sampled phyla, such as Euryarchaeota (from 54 to 76 genera), Actinobacteria (from 73 to 149), Bacteroidetes (from 55 to 107), Firmicutes (from 73 to 166), and Proteobacteria (from 265 to 538 genera; see Supplementary Table S2). To further improve the coverage of the microbial diversity, we also added representatives of some incompletely characterized lineages, mostly, at the higher taxonomic levels, such as the DPANN superphylum of archaea, with examples listed in Table 2.

Table 2.

Representatives of poorly characterized prokaryotic lineages in COGs

Organism namea	Taxonomy (phylum)	GenBank accession
ARCHAEA
Ca. Mancarchaeum acidiphilum Mia14	DPANN group, Cand. Micrarchaeota	CP019964
Ca. Nanopusillus acidilobi	DPANN group, Nanoarchaeota	CP010514
archaeon GW2011_AR15	Unclassified archaea	CP010425
BACTERIA
Armatimonadetes bacterium Uphvl-Ar1	Armatimonadetes	CP021423
Ca. Cyclonatronum proteinivorum	Balneolaeota	CP027806
Cand. division Kazan bacterium GW2011_GWA1_50_15	Cand. division Kazan-3B-28	CP011216
Cand. division SR1 bacterium Aalborg_AAW-1	Cand. division SR1	CP011268
Ca. Beckwithbacteria bacterium GW2011_GWC1_49_16	Ca. Beckwithbacteria	CP011210
Berkelbacteria bacterium GW2011_GWE1_39_12	Ca. Berkelbacteria	CP011213
Ca. Bipolaricaulis anaerobius	Ca. Bipolaricaulota	LS483254
Ca. Campbellbacteria bacterium GW2011_OD1_34_28	Ca. Campbellbacteria	CP011215
Ca. Babela massiliensis	Ca. Dependentiae	HG793133
Cand. division TM6 bacterium GW2011_GWF2_28_16	Ca. Dependentiae	CP011212
Ca. Dependentiae bacterium (ex Spumella elongata CCAP 955/1)	Ca. Dependentiae	CP025544
Ca. Gracilibacteria bacterium HOT-871	Ca. Gracilibacteria	CP017714
Ca. Peribacter riflensis	Ca. Peregrinibacteria	CP013062
Ca. Saccharibacteria bacterium YM_S32_TM7_50_20	Ca. Saccharibacteria	CP025011
Ca. Saccharibacteria oral taxon TM7x	Ca. Saccharibacteria	CP007496
Ca. Woesebacteria bacterium GW2011_GWF1_31_35	Ca. Woesebacteria	CP011214
Ca. Wolfebacteria bacterium GW2011_GWB1_47_1	Ca. Wolfebacteria	CP011209
bacterium AB1	unclassified Bacteria	CP017117
Vampirococcus sp. LiM	unclassified Bacteria	CP019384

^a- Ca. stands for Candidatus, Cand. – for Candidate.

Finally, for some bacterial and archaeal genera that were already represented in the COG database, we added selected widely studied members, primarily, either popular model organisms or important pathogens (Supplementary Table S3). For example, although the previous releases of the COGs already included Mycobacterium tuberculosis and Mycobacterium leprae, in this version, we added Mycobacterium avium, Mycobacterium marinum and Mycobacterium ulcerans. In addition, the new version includes two former Mycobacterium spp., recently transferred into novel genera: Mycolicibacterium smegmatis and Mycobacteroides abscessus (11). The genomes from Mycolicibacillus and Mycolicibacter, two other genera that have been separated from the original Mycobacterium spp. (11), did not make the list. The Streptomyces genus was originally represented by Streptomyces coelicolor, which was later replaced by the much larger genome of S. bingchenggensis. In the current release, we restored S. coelicolor and added S. griseus, which was chosen over S. avermitilis, S. cattleya and S. venezuelae as the strain used for much more research. Likewise, we added three species of Lactobacillus, two species each of Staphylococcus, Streptococcus, Clostridium and Klebsiella, as well as important pathogens such as Vibrio parahaemolyticus, Yersinia enterocolitica, Shigella flexneri andLeptospira interrogans (Supplementary Table S3). The complete list of 1309 genomes included in the current release is presented in the Supplementary Table S6 and is also available on the NCBI FTP site at https://ftp.ncbi.nih.gov/pub/COG/COG2020/data/cog-20.org.csv.

Functional annotation of COGs

After the massive overhaul of the COG annotations in the 2014 release, the current version underwent smaller changes. Nevertheless, this release includes >250 COGs with substantial updates in functional annotation and about the same number of COGs with minor name changes. In addition, annotations of approximately 60 COGs were updated to include protein domains from Pfam, InterPro or the NCBI’s Conserved Domain Database (12–14). We also compared COG annotations to those in TIGRfam and SubtiWiki databases (15,16). The notable updates in the functional annotation include enzymes involved in the modification of 16S rRNA, 23S rRNA, and tRNA (17), molybdenum cofactor biosynthesis (18), cyclic nucleotide signaling (19) and biogenesis of the cell envelope (20). A list of some of the major changes in the COG annotation introduced in the current release is given in Supplementary Table S4. The annotation process also identified several obsolete entries in the UniProt and Pfam databases, which have been reported to the curators of these databases.

Updated COG web pages

The current release introduces several changes in the setup of the COG page compared to the one in the 2014 release. In addition to the COG name and the one-letter symbol(s) for the functional category(ies) corresponding to that COG, the top line now includes four more cells. The first of these shows the short—typically, gene-based—symbol for the COG, like those used previously to denote COGs in the NCBI’s Conserved Domain Database (14). The second cell contains the name of the metabolic pathway and/or functional ensemble, if available, that the respective COG is assigned to. For the newly renamed COGs, two additional new cells contain links, respectively, to the reference(s) in PubMed that served as the basis for the new name and to a representative crystal structure in the PDB, where available.

As mentioned above, the increase in the number and diversity of the complete prokaryotic genomes occurred primarily via expansion of the already well-sampled taxa, which allowed us to retain the same arrangement of the key phyla. Based on the increased number of sequenced genomes, there are now separate sections for Deferribacteres (5 species), Negativicutes (10 species), Tissierellia (9 species), other gammaproteobacteria (6 species), and Verrucomicrobia (9 species).

In the previous releases of the COG database, gene entries in the COGs were denoted by either their GenInfo (gi) numbers or genomic locus tags in the NCBI protein database. The recent phasing out of the gi numbers (described in the NCBI Insights, see https://ncbiinsights.ncbi.nlm.nih.gov/2016/07/15/) and the shift of the entire database to GenBank protein accessions and non-unique RefSeq identifiers, coupled with the removal of numerous proteins from the database ((21,22), see https://www.ncbi.nlm.nih.gov/refseq/about/prokaryotes/faq/#FAQ5), resulted in many broken links and forced us to adopt the same identifiers for COG proteins. As a result, proteins in COGs are now listed, primarily, using their RefSeq identifiers. Proteins from 10 popular model organisms (‘Reference organisms’ in RefSeq), such as Escherichia coli str. K-12 substr. MG1655, B. subtilis subsp. subtilis str. 168, Pseudomonas aeruginosa PAO1, and Salmonella enterica subsp. enterica serovar Typhimurium str. LT2, retain their original NP_xxxxxx or YP_xxxxxxxxx identifiers that refer to a respective (single) genome. Proteins from other organisms, however, are listed under WP_xxxxxxxxx identifiers, which usually refer to many (identical) proteins from multiple genomes, sometimes even coming from diverse taxa. Accordingly, the number of protein IDs in the database is now smaller than the number of genome loci that code for these proteins. WP_xxxxxxxxx identifiers are still linked to the entries in the NCBI protein database but finding specific proteins encoded in specific genomes requires examining the ‘Identical Proteins’ link, which takes the user to the Identical Protein Groups database (https://www.ncbi.nlm.nih.gov/ipg/). The proteins that are missing in RefSeq are listed under their GenBank\ENA\DDBJ coding sequence (CDS) accession numbers.

The protein IDs are shown in full, even for multi-domain proteins where the COG covers only a part of the CDS. Accordingly, a single protein can show up in more than one COG. To determine whether the COG covers the entire protein or only a part of it, the user can simply click on the protein entry and/or check the domain organization of the respective protein via the CDD link. Mapping of the entire protein set to the COGs can be found at https://ftp.ncbi.nih.gov/pub/COG/COG2020/data/cog-20.org.csv.

Pathways and functional ensembles

The current release once again includes lists of COGs assigned to key metabolic pathways and functional assemblies. The metabolic pathways include glycolysis, pentose phosphate pathway, the TCA cycle, biosynthetic pathways for most amino acids and enzyme cofactors, biosynthesis and salvage pathways for purines and pyrimidines, and selected pathways of the lipid and murein biosynthesis. Some of these pathways are accompanied by charts prepared for the previous COG releases (23).

The functional ensembles include proteins of the large and small subunits of bacterial ribosomes, archaea-specific ribosomal proteins, enzymes catalyzing 16S rRNA, 23S rRNA, and tRNA modification and several enzymatic complexes involved in energy transformation, such as NADH dehydrogenase (Complex I), Na⁺-translocating NADH:ubiquinone oxidoreductase (NQR), Na⁺-translocating ferredoxin:NAD⁺ oxidoreductase (RNF), F_oF₁-type ATP synthase and the A/V-type ATPase. This group also includes a set of the common CRISPR-associated (Cas) proteins that was expanded from 26 to 46 COGs (see below). We believe that these functional groupings of COGs can be useful for comparative studies and plan to further expand them.

Addition of new photosynthetic and sporulation COGs

In the past, the COG approach has been used to delineate clusters of orthologous genes for cyanobacteria (24,25), lactobacilli (26), spore-forming firmicutes (27,28) and archaea (7,29,30). Making use of those data, we expanded the COG database by adding 118 COGs for proteins involved in photosynthesis and 125 COGs for sporulation proteins. Detailed analysis and annotation of cyanobacterial and firmicute COGs showed that most of them overlapped with the existing Pfam and/or TIGRfam families. However, the inclusion of these protein clusters into the COGs allowed a new outlook at the phylogenetic distribution of the respective proteins, highlighting those organisms (and lineages) that lack the respective proteins. Some of the phyletic patterns were as expected, such as the absence of the entire photosystem II in the tiny genome of the nitrogen-fixing cyanobacterium Candidatus Atelocyanobacterium thalassa (formerly cyanobacterium UCYN-A; 31). This organism was also the only cyanobacterium that lacked several poorly characterized photosynthetic proteins, including orthologs of TPR-like protein Orf03 (At3g26580 in Arabidopsis thaliana) and Acclimation of photosynthesis to environment protein APE1 (At5g38660 in A. thaliana). This analysis also confirmed the previous report that Gloeobacter violaceus PCC 7421 does not encode photosystem I proteins PsaI, PsaJ, PsaK and PsaX and photosystem II proteins PsbQ, PsbY, PsbZ and Psb27 (32). In accordance with the previous predictions (24), Candidatus Melainabacteria bacterium MEL.A1, an early branching member of the Cyanobacteria/Melainabacteria lineage, was found not to encode a single gene of photosystem I or II components (33,34).

Other phyletic patterns revealed the absence of a single gene in otherwise complete photosynthetic machinery, perhaps, as a result of sequencing or annotation problems. Indeed, photosynthetic reaction complexes include some short proteins (PsaJ, 38 aa; PsaM, 31 aa; PsbL, 37–40 aa; PsbT = Ycf8, 32 aa; Psb30 = Ycf12, 39 aa) that could be easily missed by automated genome annotation. Identification of unexpected gaps in the phyletic profiles allowed us to identify several of such missed proteins (Ycf12 in Prochlorococcus marinus, PsaJ in Anabaena cylindrica, PsbL in Fischerella sp. NIES-3754 and Planktothrix agardhii) that were absent from the protein database but are actually encoded in the respective genomes. These missing proteins were identified by tBLASTn (35) search against the respective genomic sequences, translated, and reported to RefSeq. Once included in the NCBI protein database, these genes will be added to the respective COGs.

Likewise, the 26-aa forespore membrane curvature-sensing protein SpoVM has been often overlooked in firmicute genome annotation (27,36). However, examination of the phyletic pattern of SpoVM (COG5844) showed that it was actually missing in some spore-former genomes, in keeping with its apparent non-essentiality in Clostridia (37). In accordance with previous analyses (27,38), we identified a wide variety of non-spore-forming firmicutes (mostly, among Clostridia) that nevertheless encode the master regulator of sporulation Spo0A. Examination of the phyletic patterns of Spo0A-encoding asporogens showed that most of these organisms missed at least some core sporulation genes. We expect the patterns of presence and absence of photosynthetic and sporulation genes to be useful in further studies of these important processes.

The COG approach also allowed us to create separate COGs for certain widespread domain combinations that, although clearly visible in Pfam via its ‘Domain architectures’ tool, have not been so far analyzed in any detail. As an example, the CP12 protein (COG5767) regulates the Calvin cycle in cyanobacteria and chloroplasts by forming a complex with GAPDH and phosphoribulokinase (39,40). A cyanobacteria-specific fusion of CP12 with an N-terminal tandem of CBS domains forms a separate COG5792; the members of this COG appear to regulate phosphoribulokinase in a completely different manner (41). Likewise, cyanobacteria, similarly to plants, encode multiple copies of the GUN4 ( = Ycf53) domain that was initially characterized in A. thaliana GENOMES UNCOUPLED4 mutant and has been shown to, first, regulate chlorophyll biosynthesis (by activating the Mg-chelatase), and second, to participate in plastid-to-nucleus signaling (42,43). Most proteins containing the GUN4 domain (Pfam domain PF05419) were included in COG5750. However, two widespread domain combinations of GUN4 were assigned to separate COGs. One, COG5751 that is represented in almost all cyanobacteria, consists of proteins in which the GUN4 domain is fused with an N-terminal α-helical domain (43,44). The other, COG5752, features a fusion of GUN4 with an N-terminal Ser/Thr/Tyr protein kinase domain that has been described previously (43) but never experimentally characterized. This distinct domain combination is found in up to four copies in certain cyanobacterial genomes and is, most likely, involved in an unknown signal transduction pathway.

The expanded set of Cas COGs

The recent exhaustive comparative analyses of the CRISPR-Cas systems and associated genes (45–47) allowed us to substantially expand the set of COGs that consist of Cas proteins, from 26 to 46. However, COG3512 was eliminated, and its members were moved into the other COG that includes Cas2 proteins, COG1343. The current set of Cas COGs includes all the major groups of Cas proteins, such as Cas1, Cas2, Cas3, Cas4, Cas5, Cas6, Cas7; several distinct families of Cas8; Cas10 and other proteins of the Cmr and Csm complexes of type III CRISPR-Cas systems; two major families of class 2 effectors, Cas9 and Cas12a, as well as two COGs that include CARF (CRISPR-associated Rossmann fold) domain-containing proteins, Csx1 and Csm6 (48).

USING THE COGs

Although there are several other databases of orthologous proteins in bacteria and archaea, such as eggNOG (49), KEGG Orthology (50,51), OMA (52,53), OrthoDB (54) and MBGD (55), which cover a wider selection of organisms than the COGs and provide comprehensive, automated functional annotation of their proteins, as well as specialized microbial genome databases with more narrow genome coverage (56,57), the COG database has several unique features. These include (i) classification of COGs into functional categories that allows an easy comparison of organisms based on their preference for certain types of metabolic, signal transduction, repair and other pathways; (ii) unification in a single COG of orthologous proteins from distant phylogenetic lineages, in many cases, with low (but significant) sequence similarity to each other, and (iii) phyletic patterns of presence–absence of (proteins from) the compared genomes in a given COG that facilitate functional annotation of new genomes and evolutionary inference. To our knowledge, the COG database remains the only tool that shows not only which proteins (protein families) are encoded in the given genome, but also which proteins (families) are missing in it because they are either not properly annotated (e.g. erroneously marked as pseudogenes) or not encoded in the genome at all (1,3,4,58). This makes the COGs a powerful tool for comparative and evolutionary analysis of prokaryote genomes.

As described previously (3,4,58), the most straightforward application of the COGs is the quality control of genome sequences. A set of core COGs, such as those for the key ribosomal proteins, is expected to be represented in every sequenced genome. The absence of such a COG member in a newly sequenced genome, unless it comes from an obligate intracellular parasite or symbiont with a highly reduced genome or has been lost in an entire lineage (59,60), could be a warning sign that a certain fraction of the genomic DNA is likely to be missing from the assembly. We detected several such missing ribosomal proteins, of which some simply were not properly translated, whereas others were not encoded in the available genome assemblies (see Supplementary Table S5 for examples).

Another use of the COGs involves examination of the consistency of the phyletic patterns, indicating the presence or absence of the respective protein families, among the COGs that belong to the same pathway or the same multi-subunit complex. Although not all the COGs in the same pathway or functional system are bound to have identical phyletic patterns, any deviations from the uniform pattern might indicate non-trivial evolutionary events and merit further investigation. An interesting example is the mitochondrial protein NDUFA12, a supernumerary (auxiliary) subunit A12 of the NADH:ubiquinone oxidoreductase (mitochondrial respiratory Complex I) that is required for the assembly of the complex and whose deficiency causes the Leigh syndrome (61,62). The core subunits of the NADH:ubiquinone oxidoreductase are widespread in bacteria and archaea (63), but NDUFA12 (COG3761) is represented exclusively in alphaproteobacteria, which is consistent with the alphaproteobacterial origin of the mitochondria. This gene is found in 148 of the 158 alphaproteobacterial genomes in the COGs, missing mostly in some (but not all) representatives of the family Acetobacteraceae. This subunit appears to have evolved within alphaproteobacteria and was inherited by the ancestral mitochondria, followed by the transfer of the gene into the nucleus. Meanwhile, this gene has been lost in several acetic acid bacteria that inhabit extremely nutrient-rich environments and might not require participation of NDUFA12 in the assembly of their Complex I.

Another notable example is the distribution of the c-di-AMP signaling systems. Recent analyses have identified c-di-AMP synthases and hydrolases in euryarchaea and in members of almost every bacterial phylum (19,64). However, certain members of the phylum Actinobacteria have been found to encode a typical DisA-type c-di-AMP synthase, but no (known) c-di-AMP-specific phosphodiesterase (19). If true, this would suggest unencumbered accumulation of c-di-AMP, which would be toxic for the cell (65). In short order, the missing c-di-AMP phosphodiesterase has been identified, characterized, assigned to a previously uncharacterized family (COG1524) within the alkaline phosphatase superfamily, and shown to be widespread among Actinobacteria (66,67). These examples show that analysis of identical, complementary, and mixed phylogenetic patterns can facilitate solving important biological problems.

The phyletic pattern of each specific COG can be viewed at and downloaded directly from the respective COG page, which shows the total number of organisms, genes, and proteins covered by that COG, as well as the representation of organisms from each phylogenetic group (Supplementary Figure 1). The list of proteins included in each particular COG can also be downloaded directly from the respective COG page in JSON format (in groups of 10) or extracted from the COG master file, https://ftp.ncbi.nih.gov/pub/COG/COG2020/data/cog-20.org.csv. This list of proteins can also be imported into Excel to highlight the organisms that are represented—and not represented—in the given COG. A comparison of the phyletic distribution of the COGs that belong to the same multisubunit enzyme, a metabolic pathway or a functional ensemble can be used for evaluating the phylogenetic distribution of the respective ensemble of COGs and for identifying deviations from the common pattern that could point either to problems with the annotation of certain genomes, non-orthologous gene displacement or unexpected loss of seemingly essential genes. Each of these situations merits further investigation. As a case in point, Supplementary File S2 contains an ordered list of the organisms in the current COG release (Supplementary Table S6) and a comparison (Supplementary Table S7) of the COGs representing six subunits of the Na⁺-translocating NADH:quinone oxidoreductase (Na⁺-NQR), a drug target for the recently developed new furanone antibiotic (68,69). This comparison can be used to delineate potential targets of this new antibiotic, which include such pathogens as Chlamydia pneumoniae, Chlamydia trachomatis, Klebsiella aerogenes, Neisseria gonorrhoeae, Neisseria meningitidis, Pasteurella multocida, Porphyromonas gingivalis, Vibrio cholerae, Vibrio parahaemolyticus, Yersinia enterocolitica and Yersinia pestis (Supplementary Table S7).

DATABASE CURATION AND FUTURE DIRECTIONS

The COG database is undergoing continued manual curation. COG annotations are regularly updated based on newly published functional data for the COG members. COG membership is also being updated, based, primarily, on the respective phyletic patterns. As an example, a recent examination of the phyletic patterns of the COGs for 50S and 30S ribosomal proteins has led to the identification, along with ∼500 ORFs genuinely missing in the respective genomes (59,60), of a variety of missed proteins, some with highly diverged sequences, others encoded by truncated or frameshifted ORFs, but also >70 full-size ORFs that had been overlooked in the course of genome annotation. Once these proteins are included in GenBank and/or RefSeq, they will be added to the respective COGs.

Further development of the COG database will be based, mostly, on the research projects carried out by the members of our team. Future plans include further expansion of the COG collection by adding archaeal COGs (arCOGs), splitting the COGs containing multiple paralogs, and continued refinement of COG annotations.

COG STATISTICS

The current version of COGs includes 1309 genomes from 1187 bacterial and 122 archaeal species that represent 1234 named genera, typically, with a single genome per genus. This is an almost two-fold increase from the 583 bacterial and 70 archaeal genomes that were represented in the 2014 COG release.

The database consists of 4877 COGs that include 3 236 575 unique genes mapped to 3 455 867 genomic loci, which represents 73.5% of the 4 401 819 proteins (4 110 746 bacterial and 291 073 archaeal proteins) encoded by the covered species.

DATA AVAILABILITY

The new version of the COGs is publicly available at https://www.ncbi.nlm.nih.gov/research/cog with service files available on the NCBI FTP site at https://ftp.ncbi.nih.gov/pub/COG/COG2020/. The previous versions of the COG database are available at https://ftp.ncbi.nih.gov/pub/COG/. All queries, comments, and suggestions for improvement regarding the COG database should be directed to the authors at cogs@ncbi.nlm.nih.gov.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Intramural Research Program of the U.S. National Library of Medicine at the National Institutes of Health. Funding for open access charge: Intramural Research Program of the U.S. National Library of Medicine at the National Institutes of Health.

Conflict of interest statement. None declared.

This paper is linked to: doi:10.1093/nar/gkaa1009.