Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. 1999 Apr 13; 96(8): 4285–4288.

Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles


Determining protein functions from genomic sequences is a central goal of bioinformatics. We present a method based on the assumption that proteins that function together in a pathway or structural complex are likely to evolve in a correlated fashion. During evolution, all such functionally linked proteins tend to be either preserved or eliminated in a new species. We describe this property of correlated evolution by characterizing each protein by its phylogenetic profile, a string that encodes the presence or absence of a protein in every known genome. We show that proteins having matching or similar profiles strongly tend to be functionally linked. This method of phylogenetic profiling allows us to predict the function of uncharacterized proteins.

Keywords: genomic, bioinformatics, metabolic pathways, structural complexes

The fully sequenced genomes of numerous organisms offer large amounts of information about cellular biology (see the genomes listed at the web site of The Institute for Genome Research: www.tigr.org). It is a central challenge of bioinformatics to use this information in discovering the function of proteins. Functional assignments of genes come primarily from biochemical experimentation, which can be extended by matching recently sequenced proteins to those that have already been characterized (1). For the exceptionally well studied genome of Escherichia coli (2), these and related techniques (3, 4) have lead to tentative functional assignments of slightly more than half of its proteins (5). The problem of assigning functions to the remaining proteins is addressed here.

Our computational method detects proteins that participate in a common structural complex or metabolic pathway. Proteins within these groups are defined as functionally linked. The underlying hypothesis is that functionally linked proteins evolve in a correlated fashion, and, therefore, they have homologs in the same subset of organisms. For instance, we expect to find flagellar proteins in bacteria that possess flagella but not in other organisms. In short, we show that if two proteins have homologs in the same subset of fully sequenced organisms, they are likely to be functionally linked. We exploit this property systematically to map links between all the proteins coded by a genome. In general, pairs of functionally linked proteins have no amino acid sequence similarity with each other and, therefore, cannot be linked by conventional sequence-alignment techniques.


To represent the subset of organisms that contain a homolog, we constructed a phylogenetic profile for each protein. This profile is a string with n entries, each one bit, where n corresponds to the number of genomes (16 in the present article). We indicate the presence of a homolog to a given protein in the nth genome with an entry of unity at the nth position. If no homolog is found, the entry is zero. Proteins are clustered according to the similarity of their phylogenetic profiles. Similar profiles show a correlated pattern of inheritance and, by implication, functional linkage. The method predicts that the functions of uncharacterized proteins are likely to be similar to characterized proteins within a cluster (Fig. (Fig.1).1).

Figure 1
Our method of analyzing protein phylogenetic profiles is illustrated schematically for the hypothetical case of four fully sequenced genomes (from E. coli, Saccharomyces cerevisiae, Haemophilus influenzae, and Bacillus subtilis) in which we focus on seven ...

We computed phylogenetic profiles for the 4,290 proteins encoded by the genome of E. coli by aligning (6) each protein sequence (Pi) with the proteins from 16 other fully sequenced genomes (listed at the web site of The Institute for Genome Research: www.tigr.org). Proteins coded by the nth genome are defined as including a homolog of Pi if they align to Piwith a score that is deemed statistically significant.


To test whether proteins with similar phylogenetic profiles are functionally linked, we examined the phylogenetic profiles for two proteins that are known to participate in structural complexes, the ribosome protein RL7 and the flagellar structural protein FlgL, as well as a protein known to participate in a metabolic pathway, the histidine biosynthetic protein HIS5. We first identified all other E. coli ORFs with phylogenetic profiles identical to those of these three proteins and then those ORFs with profiles that differ by one bit. The results are shown in Fig. Fig.2.2.

Figure 2
Proteins with phylogenetic profiles in the neighborhood of ribosomal protein RL7 (A), flagellar structural protein FlgL (B), and histidine biosynthetic protein His5 (C). In each case, we first found all proteins with profiles identical to our query proteins; ...

Homologs of ribosome protein RL7 are found in 10 of 11 eubacterial genomes and in yeast but not in archaeal genomes. We find that more than half of the E. coli proteins with the RL7 phylogenetic profile or profiles that differ from it by one bit have functions associated with the ribosome (Fig. (Fig.22A). Because none of these proteins have significant amino acid sequence similarity to RL7, the functional relationships to the ribosome—had they not been determined already—could not be inferred by sequence comparisons. This finding supports the idea that proteins with similar profiles are likely to belong to a common group of functionally linked proteins. Several other proteins with these profiles have no assigned function and are listed accordingly as hypothetical. The testable prediction of the clustering of phylogenetic profiles is that these as yet uncharacterized proteins have functions associated with the ribosome.

The comparisons of the phylogenetic profiles of flagellar proteins (Fig. (Fig.22B) further support the idea that proteins with similar profiles are likely to be functionally linked; 10 flagellar proteins share a common profile. Their homologs are found in a subset of five bacterial genomes: those of Aquifex aeolicus, Borrelia burgdorferi, B. subtilis, Helicobacter pylori, and Mycobacterium tuberculosis. Other proteins that appear in neighboring clusters (groups of proteins that share a common profile) include various flagellar proteins and cell-wall maintenance proteins. Flagellar and cell-wall maintenance proteins may be biochemically linked, because flagella are inserted through the cell wall. For example, the lytic murein transglycosylase MltD has a phylogenetic profile that differs by only one bit from that of the flagellar structural protein FlgL. This transglycosylase cuts the cell wall for unknown reasons. Therefore, another prediction is that this enzyme may participate in flagellar assembly.

Fig. Fig.22 A and B includes proteins in structural complexes, whereas Fig. Fig.22C shows proteins involved in amino acid metabolism. We find that more than half of the proteins with phylogenetic profiles similar (within one bit) to that of the histidine synthesis protein His5 are involved in amino acid metabolism. With the 16 currently available fully sequenced genomes, however, phylogenetic profiles are not able to separate the metabolic pathways of specific amino acids. Instead, because of the limitations of currently available data, a histidine biosynthesis protein seems to have the same profile as a tryptophan, arginine, and cysteine synthesis protein. It is probable that, as more genomes are fully sequenced and the number of entries in phylogenetic profiles is increased, similar but distinct amino acid metabolic pathways will cluster separately in phylogenetic-profile spaces.

The examples included in Fig. Fig.22 show that proteins with phylogenetic profiles similar to a query protein are likely to be functionally linked with it. We next show the converse: that groups of proteins known to be functionally linked often have similar phylogenetic profiles. As shown in Table Table1,1, we chose groups of E. coli proteins that share a common keyword in their SwissProt (7) annotation, reflecting well known families of functionally linked proteins. Because homologous proteins coded by the same genome necessarily have similar profiles, they were eliminated from the groups. For each group, we computed the number of protein pairs that are neighbors; neighbors were defined as proteins whose profiles differ by less than 3 bits. For a group of n proteins there are at most [n(n − 1)]/2 possible neighbors.

Table 1
Phylogenetic profiles link protein with similar keywords

The similarity of the phylogenetic profiles of the proteins that share a common keyword was evaluated by a statistical test; we compared the number of neighbors found in our keyword groups to the average number of neighbors found in a group of the same size but with randomly selected E. coli proteins. We found that, on average, the random sets contain very few neighbors compared with the keyword groups, even though the keyword groups contain only a fraction of all possible neighbor pairs. Thus, proteins that are functionally linked are far more likely to be neighbors in profile space than randomly selected proteins. However, we find only a fraction of all possible neighbors within a group. Therefore, not all functionally linked proteins have similar profiles; they may fall into multiple clusters in profile space. It is interesting to note that hypothetical proteins are also more likely to be neighbors than random proteins, suggesting that many hypothetical proteins are part of uncharacterized pathways or complexes.

A second indication that functionally linked proteins are likely to have similar phylogenetic profiles comes from the analysis of classes of proteins obtained from the EcoCyc library (Encyclopedia of Escherichia coli Genes and Metabolism, ref. 8). We selected several classes that contain more than 10 members and that represent well known biochemical pathways. The results of our analysis are listed in Table Table2.2. The conclusions that we draw from this analysis are similar to those found with the keyword groups: members of the group are far more likely to have neighboring profiles than members of a randomly selected control group.

Table 2
Phylogenetic profiles link proteins in EcoCyc classes

Finally, we attempted to determine the ability of our method to predict the function of uncharacterized proteins. We equate the function of a protein with that of its neighbors in phylogenetic-profile space. This equation is accomplished by means of the keyword annotations found in the SwissProt database. To test the efficacy of this method, we compared the keywords of each characterized protein to those of the neighbors in phylogenetic-profile space. All of the neighbors, in this case, were other proteins with identical profiles. We found that on average 18% of the neighbor keywords overlapped the known keywords of the query protein. By comparison, random proteins had only a 4% overlap with the same set of neighbors. We make the rough estimate that, for more than half of E. coli proteins, we can assign the general function correctly by examining the functions of their phylogenetic-profile neighbors. This estimate should also hold true for the ability of phylogenetic profiles to assign functions to uncharacterized proteins.


The phylogenetic profile of a protein describes the presence or absence of homologs in organisms. Proteins that make up multimeric structural complexes are likely to have similar profiles. Also, proteins that are known to participate in a given biochemical pathway are likely to be neighbors in the space of phylogenetic profiles. These findings indicate that comparing profiles is a useful tool for identifying the complex or pathway in which a protein participates. Finally, we were able to make functional assignments of uncharacterized proteins by examining the function of proteins with identical phylogenetic profiles.

As the number of fully sequenced genomes increases, scientists will be able to construct longer and potentially more informative protein phylogenetic profiles. There are at least 100 genome projects underway that are due to be completed within the next few months. These data will enable the construction of profiles 100 bits rather than 16 bits in length. Because the number of profile patterns grows exponentially with the number of fully sequenced genomes, the results of 100-bit comparisons should be considerably more informative than those with 16 bits. Furthermore, because the newly sequenced genomes will include several eukaryotic organisms, protein phylogenetic profiles also should become a useful tool for studying structural complexes and metabolic pathways in these higher organisms.


This work was supported by a postdoctoral fellowship from the Sloan Foundation and the Department of Energy (to M.P.), by a Hollaender postdoctoral fellowship from the Department of Energy and the Oak Ridge Institute for Science and Education (to E.M.), and by grants from the Department of Energy and the National Institutes of Health.


Encyclopedia of Escherichia coli Genes and Metabolism

Note Added in Proof

Note Added in Proof

A data structure similar to the one described in Methods has been proposed independently by Regan and Gaasterland (9) for describing the distribution of proteins in genomes.


The statistical significance of an alignment score is described by the probability (P) of obtaining a higher score when the sequences are shuffled. To compute a P value threshold, we first consider the total number of sequence comparisons that we are performing. If there are n proteins in E. coli and m in all other genomes, this number is n × m. If we were to compare this number of random sequences, we would expect one pair to yield a P value of 1/(n × m) by chance. We, therefore, set this P value as our threshold.


1. Bork P, Dandekar T, Diaz-Lazcoz Y, Eisenhaber F, Huynen M, Yuan Y. J Mol Biol. 1998;283:707–725. [PubMed]
2. Blattner F R, Plunckett G, Bloch C A, Perna N T, Burland V, Riley M, Collado-Vides J, Glasner J D, Rode C K, Mayhew G F, et al. Science. 1997;265:1453–1474. [PubMed]
3. Tatusov R L, Mushegian A R, Bork P, Brown N P, Hayes W S, Borodovsky M, Rudd K E, Koonin E V. Curr Biol. 1996;6:279–291. [PubMed]
4. Andrade M A, Sander C. Curr Opin Biotechnol. 1997;8:675–683. [PubMed]
5. Riley M. Nucleic Acids Res. 1998;26:54. [PMC free article] [PubMed]
6. Altschul S F, Madden T L, Schaffer A A, Zhang J, Zhang Z, Miller W, Lipman D J. Nucleic Acid Res. 1997;25:3389–3402. [PMC free article] [PubMed]
7. Bairoch A, Apweiler R. Nucleic Acids Res. 1998;26:38–42. [PMC free article] [PubMed]
8. Karp P, Riley M, Paley S, Pellegrini-Toole A. Nucleic Acids Res. 1998;26:50–53. [PMC free article] [PubMed]
9. Gaasterland T, Regan M A. Microb Comp Genomics. 1998;3:177–192. [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Cited in Books
    Cited in Books
    PubMed Central articles cited in books
  • Gene
    Gene links
  • GEO Profiles
    GEO Profiles
    Related GEO records
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...