Briefings in Bioinformatics, 20(4), 2019, 1063–1070

doi: 10.1093/bib/bbx117
Advance Access Publication Date: 14 September 2017
Paper

Microbial genome analysis: the COG approach

Michael Y. Galperin, David M. Kristensen, Kira S. Makarova, Yuri I. Wolf and

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

Eugene V. Koonin
Corresponding author: Eugene V. Koonin, National Institutes of Health, National Center for Biotechnology Information, National Library of Medicine,
Bethesda, MD, 20894, USA. Tel.: +1-301-435-5913; Fax: +1-301-435-7793; E-mail: koonin@ncbi.nlm.nih.gov

Abstract
For the past 20 years, the Clusters of Orthologous Genes (COG) database had been a popular tool for microbial genome anno-
tation and comparative genomics. Initially created for the purpose of evolutionary classification of protein families, the COG
have been used, apart from straightforward functional annotation of sequenced genomes, for such tasks as (i) unification of
genome annotation in groups of related organisms; (ii) identification of missing and/or undetected genes in complete micro-
bial genomes; (iii) analysis of genomic neighborhoods, in many cases allowing prediction of novel functional systems; (iv)
analysis of metabolic pathways and prediction of alternative forms of enzymes; (v) comparison of organisms by COG func-
tional categories; and (vi) prioritization of targets for structural and functional characterization. Here we review the prin-
ciples of the COG approach and discuss its key advantages and drawbacks in microbial genome analysis.

Key words: comparative genomics; genome annotation; enzyme evolution; orthologs; paralogs

Introduction processes as lineage-specific gene duplication and loss, as

well as horizontal gene transfer [7, 8]. Owing to these com-
The success of the entire genomic enterprise critically de- plexities of the evolutionary relationships among genes, the
pends on reliable genome annotation, i.e. correct identifica- COGs have become families of co-orthologous genes that em-
tion of the genes, which includes accurate determination of body one-to-many and many-to-many relationships. Hence
gene boundaries and functional annotation of the gene prod- the term ‘orthologous groups’ (of proteins) that embraces
uct(s). The Clusters of Orthologous Groups of proteins (COGs) such more complex evolutionary relationships among genes
database has been devised as a way to allow phylogenetic and simplifies the assignment of (general) functions to genes
classification of proteins from complete microbial genomes and their products. As the genomic community gradually
[1]. While the COG system has grown over the years (Figure 1), embraced the notion of co-orthologous relationships between
the goal has always been for each COG to represent a family genes [7–9], the COGs have been re-branded Clusters of
of orthologous protein-coding genes. However, when the Orthologous Genes [10].
compared genomes are separated by long evolutionary dis- During the 20 years since the inception of the COG project,
tances and possess substantially different numbers of genes, several alternative systems for orthology analysis have been de-
evolutionary relationships between these genes are not ac- veloped [11–20], some of them implementing genome-wide
curately captured by the straightforward definition of orthol- phylogenetic analysis, which, in principle, is supposed to pro-
ogy as a one-to-one relationship because of such evolutionary vide robust resolution of evolutionary relationships between

Michael Y. Galperin is a Lead Scientist at the NCBI’s (NIH) Computational Biology Branch. He uses comparative genomics to study evolution of membrane
energetics and bacterial metabolic and signaling pathways.
David M. Kristensen is an Assistant Professor at the University of Iowa’s Department of Biomedical Engineering. He uses tools of comparative genomics,
bioinformatics and systems biology to study evolution of genes in viruses and microbes.
Kira S. Makarova is a Staff Scientist at the NCBI’s Computational Biology Branch. Her area of expertise is comparative genomics and sequence analysis of
microbial genomes.
Yuri I. Wolf is a Lead Scientist at the National Center for Biotechnology Information in Bethesda, Maryland. His research is focused on quantitative aspects
of evolutionary and comparative genomics.
Eugene V. Koonin is a Senior Investigator and Leader of the Evolutionary Genomics Group at the National Center for Biotechnology Information at the
NIH. He studies various aspects of genome evolution.
Submitted: 30 May 2017; Received (in revised form): 1 August 2017
Published by Oxford University Press 2017. This work is written by US Government employees and is in the public domain in the US.

1063
 10 64 | Galperin et al.

Figure 1. Evolution of the COG system. The numbers in parentheses indicate the number of bacterial, archaeal and eukaryotic genomes, respectively, included in the
respective COG release [1–6].

orthologs and paralogs. In practice, however, such methods are paralogs among all proteins encoded in the given genome. With
computationally expensive and fraught with artifacts at differ- incomplete genomes, there always remains the obvious possi-

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

ent stages, and therefore, simpler approaches such as the COGs bility that the true ortholog of the given gene failed to make it
continue to be widely used in microbial genomics. The popular into the final assembly. Like other methods for ortholog identi-
EggNOG database (‘Evolutionary genealogy of genes: Non- fication, the COG approach relies on sequence similarity
supervised Orthologous Groups’, http://eggnog.embl.de) applies searches against selected proteomes, aimed at the identifica-
essentially the same approach as COGs to a much greater num- tion of pairwise best hits. However, instead of imposing
ber of genomes, but fully relies on automated assignment of predetermined similarity scores for delineation of likely homo-
orthologs and does not annotate the orthologous gene clusters logs, the COG approach extends the popular concept of two-way
[21, 22]. (often also called bidirectional, symmetric or reciprocal) best
Here we briefly review the key principles underlying the COG BLAST hits in each particular proteome by adding the more
approach and its applications for genome annotation and com- stringent requirement of forming a triangle, or three-way set of
parative analysis. Rather than providing a detailed description of best BLAST matches (thus forcing the mathematical property of
the COG construction methods and the resulting collections of transitivity [7, 9]) to form a new COG. Owing to the presence of
(co)orthologous gene families, our goal here is to highlight the potential paralogs from the same lineage (inparalogs [37]), the
unresolved problems in functional annotation and the possible original approach [1] only required that at least one such tri-
ways to address them. For a description of the COG database per angle be included that represented symmetrical (bidirectional)
se, the reader is referred to the previous publications [1, 2–6]. matches, with that criteria being imposed by manual supervi-
sion of groups initially constructed with an automated method.
Later, the process of detection and collapsing such obvious
Differences between COGs and other paralogs was performed by an automated method, introduced
collections of gene and protein families in the first major update of the COGs [3] and later codified in the
EdgeSearch algorithm [38–40]. Proteins from new genomes can
Functional annotation of proteins encoded in sequenced gen-
be added to the existing COGs by using the new sequences as
omes typically relies on BLASTP [23] or, more recently, HMMer
queries for an RPS-BLAST search of the collection of position-
[24] search of protein databases for the most similar sequence,
specific scoring matrices generated from COG-specific multiple
followed by a (semi-)automated transfer of the best hit annota-
sequence alignments [41]. The query is assigned to the COG that
tion to the new protein. This approach has a number of well-
yields the best score in this search. Technically, this approach is
known drawbacks [25–28]. First, if the sequence similarity is
analogous to that used to search domain databases, such as
low, there is a distinct possibility that the two proteins have dif-
InterPro and CDD, but because the COGs contain previously
ferent functions; this problem is exacerbated in cases of transi-
identified orthologs, in this case, the best hit gives a strong indi-
tive annotation of multiple proteins in this manner. Second, the
cation of orthology. A detailed discussion of other methods for
reliance on the best hit often results in a protein ending up
ortholog identification can be found e.g. in [7, 9, 42–45]. In add-
being annotated as ‘uncharacterized’ and/or ‘putative’ even
ition to sequence similarity and phylogenetic proximity, a po-
when the function of a close homolog is already known. Third,
tentially useful criterion is genomic synteny [39, 40], which,
differences in domain architectures of homologous proteins
however, in practice is typically used for manual verification of
often result in erroneous functional assignment. Given these
the existing assignments at the quality-control stage.
systematic errors, advanced approaches for functional annota-
tion of proteins increasingly rely on curated databases of pro-
tein sequences [29], such as UniProt KnowledgeBase or
Flexible similarity cutoffs
PANTHER [30, 31], and protein domains, such as Pfam, SMART
or SUPERFAMILY [32–34]. Aggregated domain databases InterPro The advantage of the triangle-based approach for orthology infer-
and CDD, which allow an easy comparison of the annotations ence is that it dispenses with artificially imposed sequence simi-
provided by various databases, often prove to be the most effi- larity cutoffs for different protein families, some of which evolve
cient tools [35, 36]. The COG approach shares some features with dramatically different rates, and permits creation of COGs
with the curated protein family databases but differs from them from proteins that span the entire range of similarity, from barely
in several important aspects. detectable to extremely high. For example, Naþ-binding c subunits
(COG0636) of Naþ-translocating ATP synthases from bacteria and
archaea have low sequence similarity and might not be recognized
Use of complete genomes as orthologs using arbitrarily high BLAST cutoffs; to further com-
A distinct feature of the COG approach is the reliance on com- plicate the annotation, archaeal protein is often referred to as sub-
plete genome (proteome) sequences, which allows relatively unit K [46]. With strict BLAST cutoffs, recognition of orthology
simple and reliable recognition of potential orthologs and becomes particularly complicated for short proteins, including
 COGs and microbial genomes | 1065

some ribosomal proteins. The COG approach also allows separ- should not allow accumulation of any intermediate that cannot
ation of closely related paralogs, such as, for example, 3-isopropyl- be further metabolized and represents a dead end: to avoid poi-
malate dehydrogenase (LeuB) and isocitrate dehydrogenase (Icd), soning the cell, such intermediate would have to be exported
members of COG0473 and COG0538, respectively, that in most into the surrounding milieu. Likewise, an intermediate in the
other databases are assigned to the same family (PF00180 in functional metabolic pathway needs to be either imported or
Pfam, SM01329 in SMART, PS00470 in PROSITE, SSF53659 in synthesized within the cell. Although the possibility of ‘distrib-
SUPERFAMILY). uted’ pathways cannot be discarded, these simple consider-
ations prove productive when COGs are superimposed on the
metabolic map to identify the intermediates that have no
Protein family granularity in COGs known enzymes to produce or metabolize them. Identification
Flexible similarity cutoffs have the built-in advantage of allowing of such gaps in pathways often suggests alternative enzymes
the COGs to be as wide or as narrow as dictated by the evolution- that can be then identified experimentally [54, 57].
ary history of a given gene family. In the above example, the LeuB/

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

Icd family is split into two COGs, which reflects the wide distribu-
tion of these enzymes among bacteria and archaea. However, this
Functional categories of genes in COGs
family also includes even two more closely related enzymes. One Another widely used feature of the COG system is the assign-
of these is tartrate dehydrogenase/decarboxylase that has been ment of all COGs to one of the 26 functional categories. These
characterized in Pseudomonas putida and Agrobacterium vitis [47, 48]. categories have evolved over time, with several of them (B, Y,
This enzyme is closely related to LeuB, still has the isopropylma- W, Z) describing functions that are found primarily in eukary-
late dehydrogenase activity and has probably evolved from LeuB otic cells. The recently added V (Defense mechanisms) and X
in the course of the adaptation of the host bacteria to life on (Mobilome) categories provide for a more detailed description of
tartrate-rich grapevine [47]. The fourth member is homoisocitrate the dynamics of bacterial and archaeal genomes. Functional
dehydrogenase AksF, which participates in the biosynthesis of the categories are assigned in accordance with the cellular roles of
methanoarchaeal coenzyme B [49]. Homoisocitrate dehydrogen- the respective COGs, so that, for example, peptide uptake sys-
ase has been described in Methanocaldococcus jannaschii, and a var- tems are included into category E (Amino acid transport and
iety of methanogenic archaea encode closely related proteins [49]. metabolism), rather than in general ‘Transport’ or other similar
At this time, there are too few tartrate dehydrogenases to form a categories. Two functional categories of uncharacterized pro-
separate COG. As for homoisocitrate dehydrogenase, LeuB and teins, R (genes with only a generic functional prediction, typic-
AksF are co-orthologs with respect to the bacterial LeuB enzymes. ally of the biochemical activity) and S (uncharacterized genes),
Accordingly, all members of this family are currently assigned to are particularly useful, as they reflect the current level of under-
the same COG0473 (LeuB) and the same arCOG01163 in archaeal standing of protein function on the proteome level and allow
COGs [10]. In the future, methanogenic homoisocitrate dehydro- tracing the progress in experimental characterization and com-
genases might form an archaea-specific COG. For now, however, putational analysis of widespread protein families. The fraction
the split of the family into two COGs appears to represent a rea- of proteins from a given genome assigned to certain COG func-
sonable compromise. In contrast, TIGRfams [50] and NCBI Protein tional categories turned out to be a useful whole-genome fea-
Clusters [51] databases divide this family into 6 and 13 clusters, re- ture [58] and has been adopted by the Genome Standards
spectively. However, because sequence similarity alone does not Consortium as an essential characteristic of the newly
allow unequivocal functional assignment, most of these clusters sequenced genomes https://standardsingenomics.biomedcen
end up with the same functional annotation, either LeuB or Icd. tral.com/submission-guidelines.

Phyletic profiles in COGs Full-length proteins and domains as COG members

An important feature of the COG approach is that a protein (or Most existing protein family databases include either full-
domain) either belongs or does not belong to it. Accordingly, a length sequences (NCBI protein database, UniProt, PANTHER,
genome is either represented in the given COG (by one or more TIGRfams [30, 31]) or separate protein domains (Pfam, SMART,
proteins) or it is not. Thus, the COG approach can dispense with SUPERFAMILY, etc [32–34]). The COG approach allows a degree
the matrix of similarity scores and replace them with the simple of flexibility: conserved domain combinations can be included
yes/no (1 or 0) representation or, alternatively, indicate the in separate COGs without the need to split them into individual
number of paralogous members of the given COG in the given domains. As an example, along with the COG0784 for individual
genome. Such phyletic patterns, i.e. the patterns of species that CheY-like receiver (REC) domains of the two-component signal
are either represented or not represented in the given COGs, are transduction systems (which also includes stand-alone CheY/
a powerful tool for functional annotation of microbial genomes Spo0F proteins), the current COG collection includes 15 add-
and evolutionary reconstruction. The most obvious use of phy- itional REC-domain COGs, such as COG2197 for DNA-binding re-
letic patterns is for identification of supposedly essential genes sponse regulators of the NarL/FixJ family, containing REC and
that are missing in certain genomes [4, 52]. Consistent applica- helix-turn-helix domains; COG0745 for DNA-binding response
tion of this principle offers an easy way to evaluate genome regulators of the OmpR/PhoB family, which consist of REC and
quality [53, 54], which is why the NCBI’s prokaryotic genome an- winged-helix domains; COG3279 for DNA-binding response
notation pipeline currently involves routine checking of the regulators of LytR/AlgR family, containing REC and LytTR do-
submitted genomes for the presence of certain (nearly) univer- mains, and many others [59, 60]. The discrimination between
sal genes, including those encoding ribosomal proteins and the architectures of proteins that share a common domain pro-
translation system components, as well as RNA polymerase vides for a finer granularity of annotation and allows better
subunits [55, 56]. A conceptually similar application of phyletic characterization of the respective proteins. However, non-
patters involves analysis of metabolic pathways and multi- critical use of COGs for high-throughput domain annotation can
protein functional systems. Obviously, metabolic pathways result in egregious errors, whereby a multidomain protein
 10 66 | Galperin et al.

receives a misleading annotation of its best COG hit that has a polymerases are represented by several paralogs which form
completely different domain architecture. The recent attempts distinct orthologous families (arCOG00328, arCOG00329,
to identify specific domain architectures and limit annotation arCOG15272 and others) within this archaeal phylum (all these
transfer to proteins with the same domain combination [36] genes are out-paralogs in Crenarchaeota). In contrast, most of
have the potential to resolve this issue. those bacteria that possess the polB gene have a single copy,
which is co-orthologous to all archaeal polB genes, so archaea
and bacteria share only one orthologous family of polB, COG0417
COG annotation
(all these genes are co-orthologs among prokaryotes with sev-
Functional annotation of COGs, including assignment of COG eral in-paralogs in archaea). Such complex relationships among
names, is based on two key principles. First, reliance on ortholo- homologous genes confound COG analysis because the defin-
gous relationships for the COG construction makes it likely, ac- ition of orthology becomes mutually dependent with the phy-
cording to the ‘orthology conjecture’, that members of each letic patterns (the definition of orthology depends on the list of
COG have equivalent functions [7] (with only rare known excep- organisms where these genes are present, which itself depends

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

tions [61]). Accordingly, experimentally characterized functions on which of the homologous genes are considered orthologs
of a single member of a given COG often can be used to assign and which are not). Several formal and informal empirical rules
the functional annotation to the entire COG. Indeed, in most have been proposed to resolve this conundrum [70]. The hier-
cases, subsequent characterization of additional COG members archical orthologous groups have been implemented in such
has confirmed the validity of the initial assignment [6]. Second, databases as EggNOG, OMA and OrthoDB [14, 22, 71].
all COG names are manually curated with the goal of creating In most of the current COG collections, all COGs are equal,
the most appropriate annotation, avoiding the common anno- and there is no hierarchical structure; only in arCOGs, an extra
tation errors [25], as well as over- and under-predictions. Thus, level of super-COGs has been introduced to combine paralogous
for those COGs whose members have two or more distinct func- COGs into higher level clusters. Although the non-hierarchical
tions, the annotations (COG names) get expanded to cover the structure of COG collections is convenient for straightforward
entire range of experimental results. In some cases, the growing genome annotation, it has substantial drawbacks. Some COGs
number of distinct paralogs justifies splitting a COG into two or include closely related proteins with similar, if not identical,
more separate COGs with higher sequence conservation and biochemical activities. In such cases, assignment of a protein to
more narrowly defined functional annotation. Many COGs, a specific COG can be taken, without justification, as an indica-
however, do not include any experimentally characterized tion that the respective organism possesses one functionality
members so that their annotation has to rely on computational but not the other. A good example is the case of glutamate and
analyses alone. In such cases, inference of a robust annotation glutamine aminoacyl tRNA-synthetases (COG0008). While most
requires careful analysis of their sequences, structures, genomic bacteria encode two paralogous enzymes that charge the Glu-
neighborhoods, phyletic patterns and other cues, which re- and Gln-specific tRNAs, archaea (as well as chlamydia, chlorobi,
quires a substantial effort that, however, often leads to interest- chloroflexi, cyanobacteria and certain members of other bacter-
ing insights [62, 63]. Such efforts are essential for increasing the ial phyla) encode only glutamate-tRNA synthetase and produce
fraction of proteins that belong to well-characterized COGs be- glutamyl-tRNA by transamidation of misacylated Glu-tRNAGln
yond the figure of 60–70% that is currently obtained for most [72]. Here, both bacterial paralogs are co-orthologs for the arch-
bacterial and archaeal genomes [6]. The overall genome cover- aeal and chlamydial enzymes, which is why they end up in a
age by COGs (including the R- and S-type COGs) has stayed single COG. Obviously, splitting COG0008 into two subCOGs
largely the same over the years and currently ranges from 65% would have been a better solution, allowing a precise character-
of the total proteomes in Chlamydiae and Planctomycetes to ization of the respective enzymes. In some cases, a COG in-
>80% in Synergistetes and Thermotogae (Figure 2). This stable cludes a small subgroup with a well-characterized function but
coverage of bacterial and archaeal genomes by COGs, despite the lack of hierarchy results in annotation of generic function
the addition of numerous new genomes, is likely to reflect the only (e.g. an ABC-type transporter).
open pangenomes of most prokaryotes [65–68] and the ex- The single-level definition of orthology can even result in
tremely rapid turnover of the poorly conserved gene class. annotations that are largely arbitrary. In some cases (e.g.
Although COG annotations typically describe protein families, COG0183, Acetyl-CoA acetyltransferase), COGs are overloaded
in the most recent release of the COG database, owing to the popu- with paralogs because it is practically impossible to track all ex-
larity of COG-based annotation, many COG names have been tant genes to distinct genes in the common ancestor. On other
modified to allow functional annotation of individual proteins [6]. occasions (COG0050, Translation elongation factor EF-Tu, and
COG5256, Translation elongation factor EF-1a), lineage-specific
COGs are created for genes that are arguably orthologous be-
Unresolved problems in the COG approach cause they are sufficiently distinct. The absence of multilevel
The wide use of COGs for microbial genome annotation and hierarchy dilutes functional annotation of the characterized
comparative analysis has illuminated several problems inher- members of the COG and weakens the evolutionary reconstruc-
ent in the COG approach that warrant a brief discussion. These tions. Developing and implementing a hierarchical framework
difficulties include, among others, the issues of COG hierarchy, is one of the most pressing problems in the COG-based ap-
inclusion of paralogs, splitting proteins into separate domains proach to gene classification and genome annotation.
and scalability of the COG approach.

Whole proteins versus protein domains

Orthologs, paralogs and xenologs: the missing hierarchy As noted above, COG construction is based on clustering of orthol-
The very definition of orthology [69] inherently depends on the ogous domains that are identified as bidirectional best hits in
group of organisms under consideration [7, 9, 37]. For example, genome-specific BLAST searches. This approach, however, is
in most members of the Crenarchaeota, the family B DNA sensitive to domain rearrangements that occurred after the
 COGs and microbial genomes | 1067

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

Figure 2. Proteome coverage by the current version of COGs. Archaeal and bacterial phyla and selected classes of Firmicutes and Proteobacteria are listed as in the lat-
est release of the COG database [6]. The orange and blue columns show the fractions of the respective proteomes covered by COGs in each taxonomic group (including
R- and S-type COGs that consist of poorly characterized or uncharacterized genes), averaged over the members of that group in the COGs (the respective numbers are
shown in parentheses). The ‘Other archaea’ group includes two genomes representing, respectively, Kor- and Nanoarchaeota; the ‘Other bacteria’ group includes mem-
bers of Deferribacteres, Nitrospirae, Verrucomicrobia and other sparsely sampled phyla, as well as representatives of several candidate phyla. The bright yellow rect-
angles on top of the archaeal columns indicate the additional coverage of the archaeal proteomes in the latest version of arCOGs [10]. The hatched rectangles indicate
the additional coverage of the archaeal and bacterial proteomes in the ATGC-COGs from the latest version of the ATGCs database [64].

divergence of the analyzed set of species from their last common number of proteins and then on a search of connected triangles
ancestor. Particularly severe problems are caused by promiscuous in clusters of reciprocal best hits that scales as O(n3) with the
domains, which can attract proteins to spurious COGs through number of proteins in the cluster [38]. Inevitably, the growth of
significant but effectively irrelevant sequence similarity to the the database outpaces the availability of the computational re-
promiscuous domains. Although this problem can be addressed sources, making regular major updates of the entire COG data-
semi-automatically, e.g. by excluding the hits that cover only a base impractical. Several divide-and-conquer strategies have
small portion of the protein sequence, precise solutions still re- been used to circumvent this major difficulty. One approach
quire manual intervention. On many occasions, conserved do- that has been implemented in several COG updates includes
main architectures allowed construction of consistent COGs that accommodating the new sequences into the existing COGs first,
were not substantially affected by the presence of a shared do- then searching for potential new COGs among the sequences
main (e.g. the widespread helix-turn-helix DNA-binding domain). that do not fit the existing ones, and then, moving some se-
Conversely, the diversity of domain architectures of proteins quences from the old COGs to the new ones [10]. The principal
involved in microbial signal transduction and containing a num- direction, however, has involved construction of dedicated COG
ber of promiscuous domains (PAS, GAF, CHASE, GGDEF, EAL and collections for distinct microbial taxa. In particular, the COGs
others) required splitting some of these proteins into individual for archaea (arCOGs) went through several closely curated re-
domains or domain combinations. As a result, the COGs are a mix leases and remain up to date, having become a widely used
of (i) highly specific domain architectures (such as the above- framework for archaeal genome annotation and analysis [10,
mentioned response regulators), (ii) multiple domain architectures 70, 73]. As illustrated in Figure 2, detailed analysis of archaeal
that include a single shared domain and (iii) separate promiscu- protein families increased the coverage of cren-, eury- and
ous domains. To our knowledge, as of this writing, there is no thaumarchaeal genomes by 18–20%, so that arCOGs now cover
complete, formal solution for optimal dissection of full-length pro- >92% of the proteins encoded in typical genomes of
teins into orthologous domains. At present, for the analysis of Crenarchaeota and Euryarchaeota. Separate projects have
multidomain proteins, the best practical approaches are offered involved construction and analysis of COGs for Cyanobacteria
by integrated domain identification tools, such as CDD (which in- and Gram-positive bacteria of the order Lactobacillales [74, 75].
cludes the COGs) and InterPro. The COG approach was also implemented in the database of
Alignable Tight Genome Clusters (ATGC) that includes closely
related bacterial and archaeal genomes [64, 76]. COGs have been
constructed separately for each ATGC. These ATGC-COGs
Scalability of the COG approach and
largely avoid the problems inherent in the COG analysis at
specialized COG collections larger evolutionary distances (lineage-specific paralogy,
The basic COG approach relies first on an exhaustive all- differential gene loss and differences in domain architectures)
against-all protein comparison that scales as O(n2) with the total and have proved an efficient platform for various types of
 10 68 | Galperin et al.

evolutionary reconstructions [77, 78]. In taxa for which ATGCs Funding

are available—i.e. those studied in sufficient depth so that mul-
tiple closely related genomes are available—the coverage of The authors are supported by Intramural Research Program
genomes is again raised so that ATGC-COGs now cover >95% of of the US National Institutes of Health at the National
the proteins encoded in typical genomes (Figure 2). Library of Medicine. D.M.K. acknowledges the support of the
The COG approach has also been extended beyond cellular Department of Biomedical Engineering at the University of
organisms to construct COG for viruses that infect bacteria or Iowa (Iowa City, USA).
archaea, and for the large DNA viruses of eukaryotes [79, 80].
The successful application of the early versions of the COGs
was to a large extent based on comprehensive manual curation
References
of the COG membership, COG names and supporting informa- 1. Tatusov RL, Koonin EV, Lipman DJ. A genomic perspective on
tion, and a substantial body of computational analysis aimed at protein families. Science 1997;278:631–7.
predicting functions for poorly characterized COGs. This effort 2. Koonin EV, Tatusov RL, Galperin MY. Beyond complete gen-

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

has led to several notable breakthroughs that have been vali- omes: from sequence to structure and function. Curr Opin
dated by subsequent experiments and opened up new research Struct Biol 1998;8:355–63.
directions, including the characterization of the CRISPR-Cas sys- 3. Tatusov RL, Galperin MY, Natale DA, et al. The COG database:
tem [81, 82], prediction of the archaeal exosome [83], identifica- a tool for genome-scale analysis of protein functions and evo-
tion of the bacterial c-di-GMP-centered signaling network [84, lution. Nucleic Acids Res 2000;28:33–6.
85], new bacterial toxin-antitoxin systems [86–88] and archaeal 4. Tatusov RL, Natale DA, Garkavtsev IV, et al. The COG data-
type IV secretion systems [89], and allowing prioritization of base: new developments in phylogenetic classification of pro-
uncharacterized proteins (COGs) for further study [90, 91]. teins from complete genomes. Nucleic Acids Res 2001;29:22–8.
However, scaling this labor-consuming approach to accommo- 5. Tatusov RL, Fedorova ND, Jackson JD, et al. The COG database:
date the exponentially growing amount of genomic sequence an updated version includes eukaryotes. BMC Bioinformatics
data is even more challenging than keeping the COGs up to 2003;4:41.
date. That path forward is likely to combine improved auto- 6. Galperin MY, Makarova KS, Wolf YI, et al. Expanded microbial
matic approaches to functional annotation with subprojects genome coverage and improved protein family annotation in
focusing on specific taxa or functional classes of COGs. the COG database. Nucleic Acids Res 2015;43:D261–9.
7. Koonin EV. Orthologs, paralogs, and evolutionary genomics.
Annu Rev Genet 2005;39:309–38.
Concluding remarks 8. Gabaldon T, Koonin EV. Functional and evolutionary implica-
The COG approach for identification of orthologous genes was de- tions of gene orthology. Nat Rev Genet 2013;14:360–6.
veloped as a platform for comparative genomic analysis shortly 9. Kristensen DM, Wolf YI, Mushegian AR, et al. Computational
after the first few microbial genomes have been sequenced. It methods for gene orthology inference. Brief Bioinform 2011;12:
could have been expected that in 20 years, this simple strategy 379–91.
based on sequence similarity hierarchy would completely give 10. Makarova KS, Wolf YI, Koonin EV. Archaeal Clusters of
way to more sophisticated, phylogenetic approaches. This, how- Orthologous Genes (arCOGs): an update and application for
ever, is not the case, primarily, because the extended orthology analysis of shared features between Thermococcales,
conjecture, according to which bidirectional best hits between Methanococcales, and Methanobacteriales. Life 2015;5:818–40.
genomes correspond to orthologs, and the latter possess equiva- 11. Kanehisa M, Sato Y, Kawashima M, et al. KEGG as a reference
lent functions, largely holds for prokaryotes given the limited ex- resource for gene and protein annotation. Nucleic Acids Res
tent of lineage-specific paralogy, differential gene loss and 2016;44:D457–62.
domain shuffling. In contrast, in eukaryotes where all these con- 12. Chen F, Mackey AJ, Stoeckert CJ, Jr, et al. OrthoMCL-DB: query-
founding aspects of genome evolution are pervasive, the COG ap- ing a comprehensive multi-species collection of ortholog
proach encounters great difficulties, and robust, genome-wide groups. Nucleic Acids Res 2006;34:D363–8.
orthology assignment does not seem to be feasible without full- 13. Uchiyama I, Mihara M, Nishide H, et al. MBGD update 2015:
scale phylogenomics. Thus, the COGs are likely to remain an im- microbial genome database for flexible ortholog analysis uti-
portant tool for microbial genome analysis for years to come, so lizing a diverse set of genomic data. Nucleic Acids Res 2015;43:
that investment of effort into refinements of this straightforward D270–6.
approach seems to be justified. 14. Altenhoff AM, Skunca N, Glover N, et al. The OMA orthology
database in 2015: function predictions, better plant support,
synteny view and other improvements. Nucleic Acids Res 2015;
Key Points 43:D240–9.
15. Heinicke S, Livstone MS, Lu C, et al. The Princeton Protein
• Robust orthology identification is essential for accurate
Orthology Database (P-POD): a comparative genomics ana-
genome annotation. lysis tool for biologists. PLoS One 2007;2:e766.
• Reconstructions of genome evolution are based on
16. Huerta-Cepas J, Capella-Gutierrez S, Pryszcz LP, et al.
orthology and paralogy. PhylomeDB v4: zooming into the plurality of evolutionary
• COGs are an essential tool in microbial genomics.
histories of a genome. Nucleic Acids Res 2014;42:D897–902.
• Several specialized COG projects have been developed.
17. Kriventseva EV, Tegenfeldt F, Petty TJ, et al. OrthoDB v8: up-
date of the hierarchical catalog of orthologs and the underly-
ing free software. Nucleic Acids Res 2015;43:D250–6.
Acknowledgments 18. Powell S, Forslund K, Szklarczyk D, et al. eggNOG v4.0: nested
The authors would like to thank all former members of the orthology inference across 3686 organisms. Nucleic Acids Res
COG team for their contributions to the project. 2014;42:D231–9.
 COGs and microbial genomes | 1069

19. Sonnhammer EL, Ostlund G. InParanoid 8: orthology analysis 40. Dewey CN. Positional orthology: putting genomic evolution-
between 273 proteomes, mostly eukaryotic. Nucleic Acids Res ary relationships into context. Brief Bioinform 2011;12:401–12.
2015;43:D234–9. 41. Marchler-Bauer A, Zheng C, Chitsaz F, et al. CDD: conserved
20. Kaduk M, Riegler C, Lemp O, et al. HieranoiDB: a database of domains and protein three-dimensional structure. Nucleic
orthologs inferred by Hieranoid. Nucleic Acids Res 2017;45: Acids Res 2013;41:D348–52.
D687–90. 42. Alexeyenko A, Tamas I, Liu G, et al. Automatic clustering of
21. Jensen LJ, Julien P, Kuhn M, et al. eggNOG: automated con- orthologs and inparalogs shared by multiple proteomes.
struction and annotation of orthologous groups of genes. Bioinformatics 2006;22:e9–15.
Nucleic Acids Res 2008;36:D250–4. 43. Chen F, Mackey AJ, Vermunt JK, et al. Assessing performance
22. Huerta-Cepas J, Szklarczyk D, Forslund K, et al. eggNOG 4.5: a of orthology detection strategies applied to eukaryotic gen-
hierarchical orthology framework with improved functional omes. PLoS One 2007;2:e383.
annotations for eukaryotic, prokaryotic and viral sequences. 44. Altenhoff AM, Dessimoz C. Phylogenetic and functional as-
Nucleic Acids Res 2016;44:D286–93. sessment of orthologs inference projects and methods. PLoS

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

23. Altschul SF, Madden TL, Schaffer AA, et al. Gapped BLAST and Comput Biol 2009;5:e1000262.
PSI-BLAST: a new generation of protein database search pro- 45. Altenhoff AM, Dessimoz C. Inferring orthology and paralogy.
grams. Nucleic Acids Res 1997;25:3389–402. Methods Mol Biol 2012;855:259–79.
24. Eddy SR. Accelerated profile HMM searches. PLoS Comput Biol 46. Mulkidjanian AY, Galperin MY, Makarova KS, et al.
2011;7:e1002195. Evolutionary primacy of sodium bioenergetics. Biol Direct
25. Galperin MY, Koonin EV. Sources of systematic error in func- 2008;3:13.
tional annotation of genomes: domain rearrangement, non- 47. Tipton PA, Beecher BS. Tartrate dehydrogenase, a new mem-
orthologous gene displacement and operon disruption. In ber of the family of metal-dependent decarboxylating R-
Silico Biol 1998;1:55–67. hydroxyacid dehydrogenases. Arch Biochem Biophys 1994;313:
26. Schnoes AM, Brown SD, Dodevski I, et al. Annotation error in 15–21.
public databases: misannotation of molecular function in en- 48. Salomone JY, Crouzet P, De Ruffray P, et al. Characterization
zyme superfamilies. PLoS Comput Biol 2009;5:e1000605. and distribution of tartrate utilization genes in the grapevine
27. Gilks WR, Audit B, De Angelis D, et al. Modeling the percola- pathogen Agrobacterium vitis. Mol Plant Microbe Interact 1996;
tion of annotation errors in a database of protein sequences. 9:401–8.
Bioinformatics 2002;18:1641–9. 49. Howell DM, Graupner M, Xu H, et al. Identification of enzymes
28. Valencia A. Automatic annotation of protein function. Curr homologous to isocitrate dehydrogenase that are involved in
Opin Struct Biol 2005;15:267–74. coenzyme B and leucine biosynthesis in methanoarchaea. J
29. Gaudet P, Livstone MS, Lewis SE, et al. Phylogenetic-based Bacteriol 2000;182:5013–16.
propagation of functional annotations within the Gene 50. Haft DH, Selengut JD, Richter RA, et al. TIGRFAMs and genome
Ontology Consortium. Brief Bioinform 2011;12:449–62. properties in 2013. Nucleic Acids Res 2013;41:D387–95.
30. Mi H, Huang X, Muruganujan A, et al. PANTHER version 11: ex- 51. Klimke W, Agarwala R, Badretdin A, et al. The national center
panded annotation data from Gene Ontology and Reactome for biotechnology information’s protein clusters database.
pathways, and data analysis tool enhancements. Nucleic Acids Nucleic Acids Res 2009;37:D216–23.
Res 2017;45:D183–9. 52. Yutin N, Puigbo P, Koonin EV, et al. Phylogenomics of prokary-
31. The UniProt Consortium. UniProt: the universal protein otic ribosomal proteins. PLoS One 2012;7:e36972.
knowledgebase. Nucleic Acids Res 2017;45:D158–69. 53. Natale DA, Galperin MY, Tatusov RL, et al. Using the COG
32. Letunic I, Doerks T, Bork P. SMART: recent updates, new de- database to improve gene recognition in complete genomes.
velopments and status in 2015. Nucleic Acids Res 2015;43: Genetica 2000;108:9–17.
D257–60. 54. Koonin EV, Galperin MY (2003) Sequence—Evolution—Function:
33. Oates ME, Stahlhacke J, Vavoulis DV, et al. The SUPERFAMILY Computational Approaches in Comparative Genomics. Boston:
1.75 database in 2014: a doubling of data. Nucleic Acids Res Kluwer Academic.
2015;43:D227–33. 55. Tatusova T, Ciufo S, Fedorov B, et al. RefSeq microbial gen-
34. Finn RD, Coggill P, Eberhardt RY, et al. The Pfam protein fami- omes database: new representation and annotation strategy.
lies database: towards a more sustainable future. Nucleic Acids Nucleic Acids Res 2014;42:D553–9.
Res 2016;44:D279–85. 56. Tatusova T, DiCuccio M, Badretdin A, et al. NCBI prokaryotic
35. Finn RD, Attwood TK, Babbitt PC, et al. InterPro in 2017-be- genome annotation pipeline. Nucleic Acids Res 2016;44:
yond protein family and domain annotations. Nucleic Acids 6614–24.
Res 2017;45:D190–9. 57. Galperin MY, Koonin EV. Functional genomics and enzyme
36. Marchler-Bauer A, Bo Y, Han L, et al. CDD/SPARCLE: functional evolution. Homologous and analogous enzymes encoded in
classification of proteins via subfamily domain architectures. microbial genomes. Genetica 1999;106:159–70.
Nucleic Acids Res 2017;45:D200–3. 58. Galperin MY, Kolker E. New metrics for comparative gen-
37. Sonnhammer EL, Koonin EV. Orthology, paralogy and pro- omics. Curr Opin Biotechnol 2006;17:440–7.
posed classification for paralog subtypes. Trends Genet 2002; 59. Galperin MY. Structural classification of bacterial response
18:619–20. regulators: diversity of output domains and domain combin-
38. Kristensen DM, Kannan L, Coleman MK, et al. A low- ations. J Bacteriol 2006;188:4169–82.
polynomial algorithm for assembling clusters of orthologous 60. Galperin MY. Diversity of structure and function of response
groups from intergenomic symmetric best matches. regulator output domains. Curr Opin Microbiol 2010;13:150–9.
Bioinformatics 2010;26:1481–7. 61. Diaz R, Vargas-Lagunas C, Villalobos MA, et al. argC orthologs
39. Lechner M, Hernandez-Rosales M, Doerr D, et al. Orthology from Rhizobiales show diverse profiles of transcriptional effi-
detection combining clustering and synteny for very large ciency and functionality in Sinorhizobium meliloti. J Bacteriol
datasets. PLoS One 2014;9:e105015. 2011;193:460–72.
 10700 | Galperin et al.

62. Prunetti L, El Yacoubi B, Schiavon CR, et al. Evidence that 77. Novichkov PS, Wolf YI, Dubchak I, et al. Trends in prokaryotic
COG0325 proteins are involved in PLP homeostasis. evolution revealed by comparison of closely related bacterial
Microbiology 2016;162:694–706. and archaeal genomes. J Bacteriol 2009;191:65–73.
63. Zallot R, Yuan Y, de Crecy-Lagard V. The Escherichia coli 78. Ran W, Kristensen DM, Koonin EV. Coupling between protein
COG1738 member YhhQ is involved in 7-cyanodeazaguanine level selection and codon usage optimization in the evolution
(preQ0) transport. Biomolecules 2017;7:12. of bacteria and archaea. MBio 2014;5:e00956-14.
64. Kristensen DM, Wolf YI, Koonin EV. ATGC database and 79. Yutin N, Colson P, Raoult D, et al. Mimiviridae: clusters of
ATGC-COGs: an updated resource for micro- and macro- orthologous genes, reconstruction of gene repertoire evolu-
evolutionary studies of prokaryotic genomes and protein tion and proposed expansion of the giant virus family. Virol J
family annotation. Nucleic Acids Res 2017;45:D210–18. 2013;10:106.
65. Tettelin H, Masignani V, Cieslewicz MJ, et al. Genome analysis 80. Grazziotin AL, Koonin EV, Kristensen DM. Prokaryotic Virus
of multiple pathogenic isolates of Streptococcus agalactiae: Orthologous Groups (pVOGs): a resource for comparative gen-
implications for the microbial “pan-genome”. Proc Natl Acad omics and protein family annotation. Nucleic Acids Res 2017;

Downloaded from https://academic.oup.com/bib/article/20/4/1063/4158183 by CNRS user on 20 October 2021

Sci USA 2005;102:13950–5. 45:D491–8.
66. Tettelin H, Riley D, Cattuto C, et al. Comparative genomics: 81. Makarova KS, Aravind L, Grishin NV, et al. A DNA repair sys-
the bacterial pan-genome. Curr Opin Microbiol 2008;11:472–7. tem specific for thermophilic Archaea and bacteria predicted
67. Puigbo P, Lobkovsky AE, Kristensen DM, et al. Genomes in tur- by genomic context analysis. Nucleic Acids Res 2002;30:482–96.
moil: quantification of genome dynamics in prokaryote 82. Makarova KS, Grishin NV, Shabalina SA, et al. A putative RNA-
supergenomes. BMC Biol 2014;12:66. interference-based immune system in prokaryotes: compu-
68. Wolf YI, Makarova KS, Lobkovsky AE, et al. Two fundamen- tational analysis of the predicted enzymatic machinery,
tally different classes of microbial genes. Nat Microbiol 2016;2: functional analogies with eukaryotic RNAi, and hypothetical
16208. mechanisms of action. Biol Direct 2006;1:7.
69. Fitch WM. Distinguishing homologous from analogous pro- 83. Koonin EV, Wolf YI, Aravind L. Prediction of the archaeal exo-
teins. Syst Zool 1970;19:99–113. some and its connections with the proteasome and the trans-
70. Makarova KS, Sorokin AV, Novichkov PS, et al. Clusters of lation and transcription machineries by a comparative-
orthologous genes for 41 archaeal genomes and implica- genomic approach. Genome Res 2001;11:240–52.
tions for evolutionary genomics of archaea. Biol Direct 2007; 84. Galperin MY, Nikolskaya AN, Koonin EV. Novel domains of
2:33. the prokaryotic two-component signal transduction systems.
71. Zdobnov EM, Tegenfeldt F, Kuznetsov D, et al. OrthoDB v9.1: FEMS Microbiol Lett 2001;203:11–21.
cataloging evolutionary and functional annotations for ani- 85. Amikam D, Galperin MY. PilZ domain is part of the bacterial
mal, fungal, plant, archaeal, bacterial and viral orthologs. c-di-GMP binding protein. Bioinformatics 2006;22:3–6.
Nucleic Acids Res 2017;45:D744–9. 86. Makarova KS, Wolf YI, Koonin EV. Comprehensive
72. Curnow AW, Hong K, Yuan R, et al. Glu-tRNAGln amidotrans- comparative-genomic analysis of type 2 toxin-antitoxin sys-
ferase: a novel heterotrimeric enzyme required for correct tems and related mobile stress response systems in prokary-
decoding of glutamine codons during translation. Proc Natl otes. Biol Direct 2009;4:19.
Acad Sci USA 1997;94:11819–26. 87. Fozo EM, Makarova KS, Shabalina SA, et al. Abundance of type
73. Wolf YI, Makarova KS, Yutin N, et al. Updated clusters of I toxin-antitoxin systems in bacteria: searches for new candi-
orthologous genes for Archaea: a complex ancestor of the dates and discovery of novel families. Nucleic Acids Res 2010;
Archaea and the byways of horizontal gene transfer. Biol 38:3743–59.
Direct 2012;7:46. 88. Makarova KS, Wolf YI, Snir S, et al. Defense islands in bacter-
74. Mulkidjanian AY, Koonin EV, Makarova KS, et al. The cyano- ial and archaeal genomes and prediction of novel defense
bacterial genome core and the origin of photosynthesis. Proc systems. J Bacteriol 2011;193:6039–56.
Natl Acad Sci USA 2006;103:13126–31. 89. Makarova KS, Koonin EV, Albers SV. Diversity and evolution
75. Makarova KS, Koonin EV. Evolutionary genomics of lactic of type IV pili systems in Archaea. Front Microbiol 2016;7:667.
acid bacteria. J Bacteriol 2007;189:1199–208. 90. Galperin MY, Koonin EV. 0 Conserved hypothetical0 proteins:
76. Novichkov PS, Ratnere I, Wolf YI, et al. ATGC: a database of prioritization of targets for experimental study. Nucleic Acids
orthologous genes from closely related prokaryotic genomes Res 2004;32:5452–63.
and a research platform for microevolution of prokaryotes. 91. Galperin MY, Koonin EV. From complete genome sequence to
0
Nucleic Acids Res 2009;37:D448–54. complete0 understanding? Trends Biotechnol 2010;28:398–406.