• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of emborepLink to Publisher's site
EMBO Rep. May 2005; 6(5): 397–399.
PMCID: PMC1299317
Science and Society
Viewpoint

Genome coverage, literally speaking

Summary

The challenge of annotating 200 genomes with 4 million publications

In late 2004, 200 complete genomes had been sequenced and made available to the research community. At the time of writing this viewpoint, that number had further risen to 221 and will have undoubtedly increased again before publication. These genomes, which represent a wide range of species from archaea to human, are a highly valuable knowledge resource for the scientific community. However, the sequencing of a full genome is just the first step in research; it must be followed by the functional characterization of genes and proteins. In this context, it is interesting to see how well represented these sequenced species are in terms of publications. We have thus obtained the number of abstracts published per species and normalized that count by the number of genes in that species to obtain a comparable measure for the number of publications per gene for all completed and published genomes. This simple measure highlights the current knowledge gap between various organisms and could further serve as a guideline for selecting genomes for sequencing projects, high-throughput functional genomics and database annotation efforts.

The 200 complete genome sequences published by December 2004 included 118 genera, 166 species and 34 additional strains for 21 species. This rate translates to a doubling time of available genome sequences of less than two years (Janssen et al, 2003a). And it remains steady: in 2003, an average of one complete genome was released per week; 47 genomes were made available in the first 44 weeks of 2004. This trend will accelerate further, as more than 1,000 genome projects are currently underway (Bernal et al, 2001).

For the 221 genomes currently available, the total number of predicted proteins is 822,114, according to the COGENT database (Janssen et al, 2003b). One of the great challenges for computational and experimental genomics is the functional characterization of the genes and proteins encoded in these genomes (Eisenberg et al, 2000), a process that should be considered as continuous (Ouzounis & Karp, 2002). To achieve this goal, it is important to rely on existing knowledge and draw from previous studies conducted on these organisms. We have analysed how well each of the currently available genomes is characterized in terms of the number of publications pertinent to the corresponding species. To achieve a reliable measure for the available knowledge per genome, we obtained the number of abstracts per species—but not strain—from Medline and divided this number by the number of predicted protein-encoding genes. The corresponding ratio, which we term the Species Knowledge Index (SKI), thus reflects our current understanding of each of these species. More detailed information on the literature coverage of sequenced organisms, tracked by COGENT, is available through our GenMed server (http://cgg.ebi.ac.uk/cgi-bin/genmed/genmed.pl).

This analysis encompasses 3,806,293 Medline abstracts corresponding to 200 genomes. On average, there are 5 abstracts per gene for the first 200 genomes in the COGENT database, ranging from 1 abstract per 1000 genes for poorly characterized species, to 55.1 abstracts per gene for Escherichia coli (Fig 1A). This arithmetic mean value is grossly distorted by a few outliers (Fig 1B; colour-coded), namely E. coli, human (48.5), Staphylococcus aureus strains (ca. 16-17), mouse (15.6), Helicobacter pylori strains (ca. 13) and Saccharomyces cerevisiae (10.6). If these outliers are excluded, then the average SKI value drops to 0.9 (580,016 abstracts for 651,183 genes). Yet, this value is still dominated either by important pathogens, such as Chlamydia trachomatis (9.5) and Haemophilus influenzae (8.9), or by model organisms such as Pseudomonas aeruginosa (5.7) and Bacillus subtilis (4.8).

...the SKI measure reflects our current status of knowledge for each species or strain at the molecular level...

Figure 1
Relationship between the number of genes and number of abstracts for the first 200 genomes made available in COGENT (v. 207, 29 Oct 2004, Janssen et al, 2003b; abstracts collected from Medline on 5 Nov 2004). (A) The x-axis represents the genome rank, ...

Other model species follow at lesser ranks, for instance Drosophila melanogaster (1.3), Schizosaccharomyces pombe (1.1) and Caenorhabditis elegans (0.5). The most studied archaeal species are Methanobacterium thermoautotrophicum and Methanococcus jannaschii, both with SKI values of 0.3. At the very end of the scale are species that have been recently characterized or are of environmental interest (Janssen et al, 2003a), such as Gloeobacter violaceus and Oceanobacillus iheyensis which have only a few abstracts, corresponding to an average of 1 per 1,000 genes. It is worth noting that there is a tendency for the SKI value to decrease in value with respect to the order in which a genome was completed (not shown).

These results demonstrate a number of limitations to our approach. First, the time of characterization of the corresponding species is not taken into account; obviously, recently characterized species have lower SKI values. A rate value that takes into account the number of abstracts per gene per year would produce a more accurate estimate for the increase in scientific interest for each organism. Second, naming conventions are not strictly followed in the literature, thus making it difficult to accurately retrieve the number of abstracts for Homo sapiens, for example. Not all abstracts that contain the word 'human' necessarily relate to molecular biology and genes, but they might contain relevant information; as a filter, we used the keyword 'protein' for human and mouse to focus on molecular information. In addition, species that are used in biotechnology, such as E. coli, are unquestionably over-represented. Similarly, there is no sufficient resolution for strain names; however, it is safe to assume that closely related strains have similar properties and the body of existing knowledge might therefore be easily transferable across strains in most cases. Third, Medline indices primarily cover medical literature and related biological sciences. It is therefore conceivable that journals that publish studies on organisms of environmental or industrial interest are not included.

Certain communities might not successfully advocate the sequencing of their favourite species, despite its relative importance in terms of available abstracts in Medline...

The accuracy of the SKI measure for organisms could further improve if the dynamics of their nomenclature were considered. For instance, Ralstonia solanacearum had 159 hits in Medline, but this organism was first classified as Burkholderia solanacearum (50 hits) and thereafter as Pseudomonas solanacearum (167 hits). The keyword 'solanacearum' produced 320 hits, close to the sum for all three species (376 hits). The slight discrepancy results from the use of more than one name in abstracts, for instance “...in Ralstonia solanacearum (formerly Pseudomonas solanacearum), a phytopathogenic bacterium...”. In this particular case, the term 'solanacearum' would thus result in a more accurate count of scientific papers for the organism R. solanacearum, assuming that no other species name contains this term. However, it would be unrealistic to consider the dynamics in nomenclature for every single species. Nevertheless, it may be possible to obtain improved SKI values in the future, assuming that the literature keeps pace with taxonomic modifications. In this respect, it may be worthwhile to use the yearly average of SKI values over a suitably long period and compare those values only with each other. In essence, global and average SKI values are not static but will change over time, thus reflecting renewed or dwindling scientific interest in a particular organism.

Despite the above limitations—namely the time of species characterization, taxonomic conventions and potential biases—the SKI measure reflects our current status of knowledge for each species or strain at the molecular level and essentially delimits the number of published studies according to the number of genes. To achieve an improvement, better descriptions of strains and species in the literature are necessary. We have recently shown that current definitions of species and genera in particular are not entirely satisfactory, according to a measure of genome sequence similarity called genome conservation (Kunin et al, 2005). It is thus imperative that sufficient literature coverage, which reflects the active experimental interest of research communities, is available for a given organism before a genome sequencing project is initiated. Certain communities might not successfully advocate the sequencing of their favourite species, despite its relative importance in terms of available abstracts in Medline, whereas closely related species with less literature support may be sequenced.

To exemplify this point, we have analysed the group of Streptomyces species that have been or are in the process of being sequenced, according to COGENT and the Genomes OnLine Database (GOLD; Bernal et al, 2001) respectively. The Streptomyces genus is a complex and important group of Actinobacteria, with many unresolved branches and a variety of phenotypic attributes (Anderson & Wellington, 2001). Two species whose genomes have been sequenced, S. coelicolor and S. avermitilis, are represented in Medline by 1014 and 119 abstracts, respectively, corresponding to SKI values of 0.13 and 0.014. According to GOLD, five other Streptomyces species are currently being sequenced: S. ambofaciens (116 abstracts), S. diversa (no abstracts), S. noursei (81 abstracts), S. peucetius (87 abstracts) and S. scabies (60 abstracts). It is therefore surprising to find that S. aureofaciens (365 abstracts), S. antibioticus (277 abstracts) and S. griseus (1223 abstracts) are not listed in GOLD as ongoing projects, although all three strains are representative species of key Streptomyces groups (Anderson & Wellington, 2001).

Obviously, the criterion of published abstracts alone is not sufficient to prioritize genome sequencing targets, yet it provides a rational measure of current knowledge and interest by taking into account the number of published studies. It is conceivable that a more elaborate listing of all species and/or strains could be obtained and ranked by the corresponding number of published abstracts. These corpora could serve as focal points for the experimental communities and would facilitate the identification of 'neglected' organisms that might be considered for genome sequencing in the future. Other uses could obviously benefit computational analysis, including database annotation and text mining.

We believe that the SKI measure demonstrates the significant variation in the number of publications for each genome and the huge challenge of using this literature to accurately annotate these genome sequences. Simply put, we would need to achieve more than 26,000 publications for Mycoplasma genitalium—which has 479 genes and is currently covered by about 400 abstracts—to reach the current SKI value for E. coli. We have a long way to go.

figure 6-7400412i1
figure 6-7400412i2
figure 6-7400412i3
figure 6-7400412i4
figure 6-7400412i5

References

  • Anderson AS, Wellington EM (2001) The taxonomy of Streptomyces and related genera. Int J Syst Evol Microbiol 51: 797–814 [PubMed]
  • Bernal A, Ear U, Kyrpides N (2001) Genomes OnLine Database (GOLD): a monitor of genome projects world-wide. Nucleic Acids Res 29: 126–127 [PMC free article] [PubMed]
  • Eisenberg D, Marcotte EM, Xenarios I, Yeates TO (2000) Protein function in the post-genomic era. Nature 405: 823–826 [PubMed]
  • Janssen P et al. (2003a) Beyond 100 genomes. Genome Biol 4: 402. [PMC free article] [PubMed]
  • Janssen P, Enright AJ, Audit B, Cases I, Goldovsky L, Harte N, Kunin V, Ouzounis CA (2003b) COmplete GENome Tracking (COGENT): a flexible data environment for computational genomics. Bioinformatics 19: 1451–1452 [PubMed]
  • Kunin V, Ahren D, Goldovsky L, Janssen P, Ouzounis CA (2005) Measuring genome conservation across taxa: divided strains and united kingdoms. Nucleic Acids Res 33: 616–621 [PMC free article] [PubMed]
  • Ouzounis CA, Karp PD (2002) The past, present and future of genome-wide re-annotation. Genome Biol 3: COMMENT2001 [PMC free article] [PubMed]

Articles from EMBO Reports are provided here courtesy of The European Molecular Biology Organization
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...