Info

# of taxa in which protein occurs

■ Two-hybrid A Non-two hybrid

Figure 5. The vertical axis shows the average degree (+ one s.d.) of proteins in the yeast protein interaction network as a function of the number of genomes—among six fully sequenced genomes—in which these proteins contain homologues, as shown on the horizontal axis. The analysis is based on two different data sets on yeast protein interactions, one ('two hybrid') using the yeast-two hybrid assay to identify such interactions,38 the other ('non-two hybrid') a publicly available database on protein interactions from which I eliminated all data generated with the two-hybrid assay.39 Protein comparisons are based on the following six maximally diverse fully sequenced and publicly available genomes: Schizosaccharomyces pombe (www.sanger.ac.uk), Plasmodium falciparum (www.plasmodb.org), Arabidopsis thaliana (www.tigr.org), Drosophila melanogaster (www.fruitfly.org), Escherichia coli K12-MG1655 (www.tigr.org), Methanococcus janaschii DSM2661 (www.tigr.org). For the data shown, I used gapped BLAST37 with a threshold protein alignment score of E < 10"5 to identify homology. Results (not shown) are qualitatively identical for threshold scores of E < 10"2 and E < 1010.

metazoa, plants, protists, fungi (exclusive Saccharomyces spp.), eubacteria, and archaea. Table 1 summarizes the results. Seven out of 15 highly connected proteins and six out of 15 proteins with degree one have homologues in all eukaryotes. The same proportion (12 out of 15) of highly connected proteins and proteins with degree one have homologues in fungi outside the genus Saccharomyces. The same holds also for proteins that have no homologues outside this genus (3 out of 15 proteins). Based on this data, it appears that highly connected yeast proteins are not phylogenetically older than proteins of low degree.

While this finding is at first sight puzzling, the following analysis suggests a mundane explanation. This explanation emerges from a stochastic model of how the number of a proteins interaction partners changes over time. Consider one protein in a protein interaction network and denote as Dt the number of proteins this protein interacts with. If time t is measured in suitable discrete units, such as million years, then the change of this variable over time can be represented by a first order Markov process.40 Specifically, designate as pi the probability that the protein gains an interaction, that is, that its degree increases by one (through a mutation that has become fixed in a population). Formally pi = Prob (D, = i+1 \Dt.j = i). Similarly, denote

Table 1. Taxonomic distribution of proteins with different connectivity in the yeast protein interaction network

High Degree Proteins

Table 1. Taxonomic distribution of proteins with different connectivity in the yeast protein interaction network

High Degree Proteins

Name

Deg

0 0

Post a comment