Our knowledge about DNA and protein sequences is far more advanced than our  knowledge about the resulting biological and biochemical functions. This is due to the rapid progress in genome sequencing: 81 genomes are already completely sequenced and 437 genome projects are in progress (Mar. 2002). Therefore, researchers are frequently forced to assign biological functions on the basis of  sequence homology alone (inferring a similar biological function from similarities in the sequences). As the sequence databases grow larger, an increasing number of assignments are based on homology to sequences whose functions been assigned only tentatively, based on homology to still other sequences.

For example, the biological function of a sequence in organism A is known. The first inference is that if organism B has a similar sequence then this sequence codes for a similar biological function. But then, when organism C has a sequence that is more similar to B than to A, researchers sometimes assume, that the sequence of C still codes for the function of A. Such an assumption can be completely wrong, as shown below.

A good strategy to avoid such pitfalls is the use of phylogenetic trees. The construction of phylogenetic trees provides information about a protein of interest in terms of its relationship to other proteins and may allow to draw conclusions about its biological functions that would not otherwise be apparent.

The term "phylogenomics" was coined by Jonathan A. Eisen [1]; he postulated that evolutionary analysis of genes improves functional predictions for uncharacterized proteins. The basis of this idea is simple: Because the functions of genes or proteins change over time as a result of evolution, the reconstruction of the evolution of genes should faciliate functional predictions for proteins with unknown functions.

Even if the function of a protein is known, there are some distinct advantages when a phylogenomic analysis during annotation of genes is performed:
 

• Error detection:
Phylogenomic methods can not only detect errors associated with the annotation, but also false a interpretation of genomic sequences. For example, during annotation of the Streptococcus pyogenes genome,it was stated that S. pyogenes possesses a gene similar to rpoE, a heat shock sigma factor of Escherichia coli(2). However, a phylogenomic analysis showed that the rpoE gene of S. pyogenes encodes the delta subunit of RNA polymerase (3).
 
 
• Visualisation:
Phylogenetic methods allow the visualisation and definition of large protein families as well as categorization of subfamilies.
 
• Speciation:
Conclusions about the origin of orthologous proteins (those with the same function in different organisms) and of paralogous proteins (those with different functions in the same organism) can be obtained by the application of phylogenetic methods. As a consequence, examples of convergent evolution can be detected.

• Taxonomy:
A gene tree, showing the relatedness between a set of similar genes, can be compared to a species tree, which displays the relatedness between a set of species.

 
• Selectivity:
Phylogenetic trees can be constructed for certain regions only: This "masking" is used to exclude regions without significance for the investigations. For example, a great deletion in a gene of a certain species would affect the similarity (in %) to homologous genes in other organisms. This deletion does not necessarily influence the functionality of the gene product. A phylogenomic analysis of a dataset could exclude the region of the deletion. Masking allows exclusion of regions that do not exhibit a biological signal. But masking can be also used to analyse and compare individual protein domains.

• Robustness:
If the mutation rate between species differs significantly, phylogenetic methods are most appropriate to establish evolutionary relationships. For example, Mycoplasma species share a common origin with gram positive bacteria. However, these organisms exhibit a higher mutation rate relative to other bacteria. As a consequence, this relatedness is however not evident, when the similarity (in %) between mycoplasmal genes and their bacterial homologs is calculated. 

(1) Eisen, J.A. 1998. Genome Res. 8: 163-167.
(2) Ferretti, J.J., et al. 2001. PNAS 98: 4658-4653.
(3) Mittenhuber, G. 2002. J. Mol. Microbiol. Biotechnol. 4: 77-91.
(4) THE GENOME INTERNATIONAL SEQUENCING CONSORTIUM. 2001. Nature 409: 860-921.
(5) Stanhope, M.J., et al. 2001. Nature 411: 940-944.

Previous  Next