Background
The introduction of small subunit ribosomal RNA as a tool in microbial taxonomy by Carl Woese and George Fox 1 led most microbiologists to assume that the concepts of animal and plant taxonomy could be extended to the realm of prokaryotes. In particular, it was assumed that a natural taxonomic system for microorganisms was feasible 2. The goal of a natural taxonomic system is the formation of taxonomic groups that are defined by shared ancestry 3. By definition, an ancestor that defines a monophyletic group can only give rise to members of this group. No organism outside this group has a lineage that traces back to the same ancestor (paraphyletic group); however, there might be earlier ancestors that define more inclusive monophyletic groups. The metaphor for organismal evolution that underlies a natural taxonomic system is a strictly bifurcating tree of species. A decade ago ribosomal RNA promised that one day it might be possible to place every extant organism on a universal tree of life, and the hope was that more genomic sequences would make this placement more accurate.
However, the analyses of completely sequenced genomes initiated a reassessment of concepts in microbial evolution 4. While some molecular markers were found to agree with one another e.g., 5, others do not 6789101112. Transfer of genetic information between divergent organisms has turned the tree of life into a net or web 13, and genomes into mosaics. Different parts of genomes have different histories, and representing the history of genome evolution as a single tree appears inconsistent with the data. Nevertheless, the assumption of a tree-like process still underlies many approaches. Genome content trees have been calculated based on the presence and absence of genes 141516 or types of protein folds 17. While there is limited agreement between genome and rRNA phylogeny, at present it remains unclear whether this similarity is based on shared ancestry of part of a less frequently exchanging genome core 18, or if the apparent congruence is itself the result of horizontal gene transfer 19.
Overall genome content is not best represented on a single tree. Fig. 1 gives an example of an alternative depiction, where thickness of a line reflects percentage of genes shared between two genomes. The coherence among the three domains of life (Bacteria, Archaea, Eucarya 20), is clearly reflected in genome content; i.e., Archaea share more genes with other Archaea than with Bacteria, but many features are incompatible with representing the relationships between different genomes as a tree. For example, the mesophilic euryarchaeon Halobacterium sp. has more genes in common with the mesophilic Bacteria than does the thermophilic crenarchaeote Aeropyrum pernix. However, the extremophilic euryarcheote Archaeoglobus fulgidus shares many more genes with the extremophilic bacteria, Aquifex aeolicus and Thermotoga maritima than does Halobacterium. While this example illustrates the web-like relationships among genomes, recent phylogenetic reconstructions from molecular data have explored only few alternatives to the tree-paradigm (e.g. 2122).
<p>Figure 1</p>
Star Like Representation of Genome Relationships.
Star Like Representation of Genome Relationships. The diagram depicts pairwise comparisons among thirteen genomes. Every genome is represented as a point on the perimeter of a circle. The thickness of the line connecting two genomes reflects the percentage of shared genes between the genomes. The thickest line connecting Aquifex aeolicus and Thermotoga maritima corresponds to 51% shared genes, and the thinnest line connecting Aeropyrum pernix and Borrelia burgdorferi corresponds to 9% shared genes. A gene is considered shared when it had a BLAST hit in the other genome with an E-value below 10-8. The percentage of genes shared between genomes A and B is calculated as ((#of genes in A shared with B/total # of genes in A)+(# of genes in B shared with A/total # of genes in B))/2. Bacteria are depicted in green, Archaea in red and Eukaryotes in blue. The domain affiliation is also indicated by a letter following the species name (A: Archaea, B: Bacteria, and E: Eukaryotes).
![]()
One obvious drawback of the star-like representation in Fig. 1 is that it utilizes BLAST search results only. Any phylogenetic information retained in the sequences is not utilized beyond the presence or absence decision based on a single expectation value cut-off. Because of recombination, individual genes themselves might be mosaic 23; however, within-gene recombination of protein coding genes occurs mostly between closely related organisms. The redundancy of the genetic code greatly reduces recombination between divergent proteins. Even if a region is 100% conserved on the amino acid level, the encoding DNA can be so different as to allow the mismatch repair system to prevent recombination. For studies of single divergent orthologous protein encoding genes the assumption of a tree-like evolutionary history remains a reasonable expectation. In this manuscript we focus on methods that utilize the phylogenetic information that is retained in molecular sequence data, while not presuming that genomes as a whole evolved in a tree-like fashion.
In an elegant approach Korbinian Strimmer and Arndt von Haeseler 24 utilized Bayesian posterior probabilities to assess the phylogenetic information contained in an alignment of four homologous sequences. With four sequences there are only three possible tree topologies, and thus the three posterior probabilities corresponding to these three trees must sum up to one. Utilizing a barycentric coordinate system, the resulting probability vector is represented as a point in an equilateral triangle (Fig. 2), where the distances of the point P to the three sides represent the three probabilities. Strimmer and von Haeseler applied this approach to depict the phylogenetic information content present in a multiple sequence alignment. They plot the results from the analyses of all possible quartets, where the four sequences are selected from a single multiple sequence alignment in the same coordinate system. If there is a lot of phylogenetic information in the alignment, then most probability vectors will fall close to one of the corners; conversely datasets containing little phylogenetic information will mainly result in vectors falling into the center of the triangle. Here we explore the application of this and similar approaches in comparative genome analyses. In particular, we compare different approaches to calculate Bayesian posterior probabilities, and we compare these probabilities to the more widely used bootstrap support values. We assess the reliability of the different probability mapping approaches through their application to mitochondrial genomes, and we illustrate their usefulness by mapping selected interphylum and interdomain relationships.
<p>Figure 2</p>
Mapping of the probability vector onto an equilateral triangle.
Mapping of the probability vector onto an equilateral triangle. Each QuartOP is represented as a probability vector P inside an equilateral triangle. The position of P is determined by the barycentric coordinates (p1, p2, p3), which correspond to the posterior probabilities or bootstrap support values of the three possible tree topologies. The vertices of the triangle T1, T2 and T3 represent the three possible unrooted tree topologies. Geometrically, each of the coordinates (p1, p2, p3) equals the distance between P and the side of the triangle opposite the corresponding vertex. Points closer to a vertex Ti have a larger corresponding probability pi and represent a more probable tree topology than the two alternatives. All the points are classified by their position in one of three zones: "total" zone, "90%" zone and "99%" zone, which are depicted schematically and not drawn to scale. In this diagram, point P corresponds to a dataset which has highest probability for the topology T3, but the probability is below 90%, so the point P is located in the "total" zone, but not in the 99% or 90% zone. Figure adapted from 24.
![]()
Results and Discussion
Overview of data flow in probability mapping
An outline of our approach to genome probability mapping is given in Figure 3. Using SEALS 25 and MySQL we developed scripts that identify and retrieve quartets of orthologous protein-encoding open reading frames (QuartOPs) from four selected genomes. We use the term genome to denote the collection of all ORFs identified in a genome. (In the case of genomes that are not well annotated, it is feasible to use a very wide definition of ORF, e.g, all amino acid sequences encoded between two stop codons in any of the six possible reading frames. As long as one of the genomes included in the analyses is properly annotated, only those identified ORFs that are actually homologous to an identified ORF will become part of a quartet of orthologs.) We utilize an operational definition of an ortholog: two open reading frames are considered orthologous, if and only if they are each other's top scoring BLAST hit when one is used as a query to search the other genome. A QuartOP is formed when each of the open reading frames picks the other members of the quartet as the top scoring hit in searches of the respective genomes. QuartOPs are similar to the clusters of orthologous groups (COGs) maintained by the NCBI 262728, but differ in that COGs require only unidirectional, circular best hit relationships for three of the reference genomes, whereas we require the reciprocal top hit relationship for the four genomes included in a quartet, and we do not limit our identification of QuartOPs to a number of reference genomes. Montague and Hutchison utilized a comparable approach in their definition of congruent COGs 29. So far we have analyzed 68 genome quartets (see supplementary material). The number of QuartOPs identified per genome quartet ranges from 82 (for genome quartet #6: Deinococcus radiodurans, Treponema pallidum, Escherichia coli, and Halobacterium sp.) to 1182 (for genome quartet #63: Agrobacterium tumifaciens, Sinorhizobium meliloti, Mezorhizobium loti and Caulobacter crescentus).
<p>Figure 3</p>
Data flow for the genome quartet analysis.
Data flow for the genome quartet analysis. See Materials and Methods for details.
![]()
Each of the aligned QuartOPs from a genome quartet was analyzed with respect to the posterior probability of the three possible tree topologies given the aligned QuartOP. Routinely we calculated these probabilities using Strimmer's and von Haeseler's approach 24: Using each of the three topologies as a usertree, we calculated the maximum likelihood for each of the three topologies given the data. We then use the three maximum likelihoods to calculate the probability for topology i according to the formula: Pi = Li/(L1+L2+L3), where Li is the likelihood for the best tree given topology i. Other types of reliability measures used to evaluate QuartOPs were bootstrap support values and Bayesian posterior probabilities estimated using MrBayes program (see below).
An example for the comparison of four genomes from different phyla is given in Fig. 4A. Surprisingly, each of the three tree topologies is strongly supported by more than 40 QuartOPs, and most of the QuartOPs appear to strongly support one of the trees. None of the three possibilities has majority support. Figure 5 lists the functional categories of those QuartOPs that strongly support the different tree topologies. None of the categories shows a preference for a particular tree topology. For each tree topology more than 50% of the strongly supporting QuartOPs belong to the category "information storage and processing", while this category contains only about 1/3 of the genes present in the genomes. While the genes in this category appear more conserved and phylogenetically informative, the strong support that the genes in this category provide is nearly evenly split between the three possibilities.
<p>Figure 4</p>
Maps of a genome quartet with organisms from four different bacterial phyla: Escherichia coli (Gram negative), Deinococcus radiodurans (Deinococcales), Bacillus subtilis (Gram positive) and Treponema pallidum (spirochete).
Maps of a genome quartet with organisms from four different bacterial phyla: Escherichia coli (Gram negative), Deinococcus radiodurans (Deinococcales), Bacillus subtilis (Gram positive) and Treponema pallidum (spirochete). Tree topologies assigned to the vertices are depicted in New Hampshire tree format near the corresponding vertex of the triangle and they are equivalent to the unrooted tree topologies as depicted in Figure 2. The three numbers associated with each tree topology indicate how many QuartOPs fall into each of the three zones: "total", 90% and 99% respectively. For definition of zones see figure 2. A) Probabilities are calculated according to Strimmer and von Haeseler 24. There is no single topology that is supported by the majority of the QuartOPs and all three possible tree topologies are supported by roughly equal number of QuartOPs at the different probability levels. B) Probabilities are calculated with MrBayes program 31. C) Bootstrap support values are plotted. For this case the zones are "total", 70% and 90% support, respectively. Bootstrapping appears to provide a more conservative reliability estimate than the posterior probabilities used in cases A and B. Nevertheless, each tree topology is still supported by a roughly equal number of bootstrapped datasets.
![]()
<p>Figure 5</p>
Distribution among different functional categories for those datasets that support one of the three topologies with better than 99% posterior probability.
Distribution among different functional categories for those datasets that support one of the three topologies with better than 99% posterior probability. Tree topologies are indicated by column numbers 1, 2 and 3. Column 1 corresponds to topology ((1,4),2,3), columns 2 and 3 correspond to topologies ((1,3),2,4) and ((1,2),3,4) respectively. Divisions into functional categories are adopted from the COG database 27. Functional categories are aggregated into four broad functional meta-categories. Distributions of datasets among the meta-categories are plotted as pie charts for each tree topology. In this case all three topologies are supported by roughly equal number of datasets from each meta-category.
![]()
Impact of model parameters and sequence conservation
To test if ill-aligned sequences might have had an impact on the analyses, we repeated the analysis of genome quartet # 8 (see Figure 4) using only QuartOPs that contained very similar sequences. By default we only excluded top hits with an E-value larger than 10-4. We repeated the example given in Fig. 4A with a cut-off of 10-20, i.e., we not only required the sequences in a QuartOP to be each others top hit, but in addition we asked for a high similarity between the two sequences. As a result the support for the three topologies in genome quartet #8 dropped to 54 (44, 38), 51 (45, 32), 39 (29, 28) (the numbers in parenthesis are the numbers of quartets that support the topology with posterior probability larger than 90% and 99%, respectively). To access the level of sequence conservation within the QuartOPs' sequences, we calculated the average percentage of pairwise identity per QuartOP. It varied from 40.53 10.54% to 43.84 ± 9.7% when the E-value cut-off was varied between 10-2 and 10-20 (see supplementary material for the summary table). While pairwise sequence identity is not a universally dependable measure of phylogenetic information content, these values illustrates that the sequences within a QuartOP are neither identical to one another, nor so divergent as to be saturated with substitutions and of questionable homology 30. Using only the most conserved QuartOPs does not change the qualitative result: each of the three possible tree topologies is supported by about an equal number of QuartOPs (see supplementary material).
We recalculated the likelihoods for all QuartOPs in genome quartet #8 using a model that incorporates among site rate variation (ASRV). The posterior probabilities calculated according to Strimmer and von Haeseler did not change dramatically and each of the three tree topologies is still supported by roughly equal number of QuartOPs. The maps for this analysis are available in the supplementary material.
Estimating Bayesian Posterior Probabilities
The formula used by Strimmer and von Haeseler 24 to calculate posterior probabilities (i.e. the probability that tree topology Ti is true given an aligned set of four sequences) considers only three trees (i.e. branch lengths and topology), each with the same prior probability. These three trees are those that have the highest likelihood for the three possible topologies. However, there are infinitely many other trees that differ from the three chosen ones only by differences in branch lengths. What is the effect on the calculated posterior probability of using only the single best tree as a representative of all the trees with the same topology? There is no a priori reason to exclude the other trees that have slightly lower likelihoods.
A different approach that does not make these assumptions is the use of Markov Chain Monte Carlo methods to explore tree space. We used the program MrBayes written by Huelsenbeck and Ronquist 31. Using a QuartOP with posterior probabilities of .76, .10 and .13 we explored different parameter choices for the biased random walk through tree-space. We chose two chains with 5,000 burn-in cycles, and 25,000 cycles with sampling after every cycle as a compromise between increased precision of the probability estimate and computation time (see Materials and Methods for more details).
The result of calculating the posterior probabilities of all QuartOPs in genome quartet #8 is given in figure 4B. Again all of the three tree topologies are strongly supported by some QuartOPs. When we repeated this analysis using the same settings, none of the probabilities changed by more than a few percent. The support for the three tree topologies changed from 67/47/37, 69/40/22 and 86/55/32 (the three numbers indicate total support and QuartOPs that supported a topology with more than .90 and .99 respectively) in the first run to 67/47/37, 70/41/22 and 85/55/32 in the second, indicating that the chosen parameters provided satisfactory reproducibility. Plots of both analyses are available in the supplementary material. Comparing figure 4B with 4A it is clear that in this case the Bayesian posterior probabilities estimated with MrBayes are more conservative assessments of reliability than the ones calculated according to 24. The 99% support level calculated according to 24 approximately corresponds to the 90% support level calculated with MrBayes.
Bootstrap support values versus posterior probabilities
To facilitate comparison of Bayesian posterior probabilities with a more widely used confidence measure, we generated 100 bootstrapped samples 32 from each QuartOP in case #8. Each of the bootstrapped samples was analyzed using maximum likelihood with the same model of substitution as before. Each of the bootstrapped samples supports one of the three possible topologies, thus the sum of the bootstrap support values for the three topologies adds up to 100%, and the percentage of bootstrapped samples for each QuartOP that best supported each tree was again plotted in a barycentric coordinate system (Fig. 4C). Many more QuartOPs map into the central region of the triangle as compared to Figure 4A and 4B. Clearly, for this test bootstrap support values are more conservative measures of support than either of the posterior probabilities calculated above. Nevertheless, there are still several QuartOPs that strongly support each of the three tree topologies; however, there are 22 QuartOPs that support grouping E. coli with Treponema pallidum with better than 90% bootstrap support, whereas the alternatives are supported by only 12 and 13 QuartOPs, respectively. Comparing Figures 4A and 4C it appears that in analyzing quartets 70% bootstrap support is comparable to .99 posterior probability calculated according to 24.
Comparison of the different reliability assessment tools
ML-mapping according to 24 is the least conservative of the tools explored. For the test cases analyzed a posterior probability of .99 according to 24, corresponds to a Bayesian posterior probability of .90 calculating using a Markov chain exploration of tree-space using 31 and about 70% bootstrap support. We did not find a strong dependence of the results on the substitution models used in calculating likelihoods and separate runs indicated satisfactory precision of the calculated probabilities and bootstrap values. Given that we only analyzed about 300 QuartOPs using all three approaches it would be premature to generalize our findings; however, other analyses that utilized both bootstrapping and Bayesian posterior probabilities also found bootstrapping to be more conservative than posterior probabilities calculated using Bayesian methods with Markov chain Monte Carlo sampling (e.g., 333435).
Mitochondrial genomes
While gene transfer into the mitochondrial genomes has been inferred 36373839, mitochondrial genomes are expected to have undergone many fewer legitimate and illegitimate recombination events than free-living prokaryotes. Clearly, if probability mapping is to be considered a reliable approach, we expect that when analyzing quartets of mitochondrial genomes, the different genes should all support the same tree topology.
In most instances, this expectation is fulfilled (see Table 1), even though we selected instances in which the splits could be expected to be ill resolved, e.g., echinoderm, mammal, insect, mollusk (m4), or protist, fungus, animal, plant (m7). The only exception was an ORF in quartet m7 that encodes the cytochrome oxidase subunit II. This ORF did not support grouping the animal with the fungal homolog as expected; rather it grouped the protist and animal homologs together (posterior probability according to 24 was » 0.99). Inspection of the aligned sequences (Fig. 6) revealed that there are more residues shared between the homologs from Cafeteria roenbergensis and Homo sapiens thanbetween the homologs from Cafeteria roenbergensis and Arabidopsis thaliana. No artifact that could be responsible for this unexpected grouping was detected. The same high support for this unexpected grouping is also recovered in bootstrap analysis and in posterior probabilities calculated with MrBayes 31.
<p>Figure 6</p>
Alignment of mitochondrial cytochrome oxidase subunit II.
Alignment of mitochondrial cytochrome oxidase subunit II. The alignment for the control mitochondrial quartet m7 (see Table 1) that supports the unexpected ((Homo sapiens, Cafeteria), Saccharomyces, Arabidopsis) topology. The exact matches for each tree topology are colored in three different colors. Blue corresponds to the ((Homo sapiens, Cafeteria), Saccharomyces, Arabidopsis), yellow corresponds to the ((Homo sapiens, Arabidopsis), Saccharomyces, Cafeteria) and green corresponds to the ((Homo sapiens, Saccharomyces), Arabidopsis, Cafeteria) tree topology. As can be seen, the majority of the matches are in favor of ((Homo sapiens, Cafeteria), Saccharomyces, Arabidopsis) tree topology. There are nine parsimony informative positions favoring the latter topology, and only three for each of the other two topologies.
![]()
<p>Table 1</p>
Results of analyses for the control mitochondrial genome quartets #m1-m7.
#
Genome 1
Genome 2
Genome 3
Genome 4
((1,2),3,4)
((1,3),2,4)
((1,4),2,3)
Tot.
A
B
Tot.
A
B
Tot.
A
B
m1
Drosophila
Drosophila
Ceratitis capitata
Apis mellifera
9
8
8
0
0
0
1
0
0
melanogaster
yakuba
ligustica
10
8
8
0
0
0
0
0
0
9
7
5
0
0
0
1
1
0
m2
Alligator
Opossum
Stork
Donkey
0
0
0
11
11
11
0
0
0
0
0
0
11
11
11
0
0
0
0
0
0
11
11
11
0
0
0
m3
Turtle
Opossum
Stork
Donkey
0
0
0
12
12
12
0
0
0
0
0
0
12
12
12
0
0
0
0
0
0
12
12
12
0
0
0
m4
Starfish
Donkey
Fruit Fly
Doorsnail
8
7
7
2
1
1
0
0
0
8
7
7
1
0
0
1
0
0
8
7
5
1
1
0
1
0
0
m6
Reclinomonas
Saccharomyces
Arabidopsis
Homo sapiens
0
0
0
4
3
3
0
0
0
americana
cerevisiae
thaliana
1
0
0
3
3
3
0
0
0
0
0
0
4
3
2
0
0
0
m7
Cafeteria
Saccharomyces
Arabidopsis
Homo sapiens
0
0
0
3
2
2
1
1
1
roenbergensis
cerevisiae
thaliana
1
0
0
2
2
1
1
1
1
1
0
0
2
1
0
1
1
1
The groupings corresponding to the expected organismal phylogenies are given in bold. The three numbers in each table cell correspond to the three approaches used. The top number corresponds to results obtained using Strimmer and von Haeseler's approach [24], the middle number corresponds to results obtained using MrBayes program [31], and the bottom number corresponds to results of bootstrap support values calculation. Column "Tot." lists the number of QuartOPs from "total" zone, column A lists the number of QuartOPs from "90%" zone (70% for bootstrap support), and column B lists the number of QuartOPs from "99%" zone (90% for bootstrap support). For definition of zones see Fig. 2. With the exception of the one dataset for quartet #m7, the analyses proved to be consistent with organismal tree topologies. The alignment for the exceptional dataset is presented in Fig. 6. The common names for the organisms listed correspond to the following scientific names: alligator corresponds to Alligator mississippiensis, opossum to Didelphis virginiana, stork to Ciconia ciconia, donkey to Equus asinus, turtle to Chelonia mydas, starfish to Asterina pectinifera, doorsnail to Albinaria caerulia, fruit fly to Drosophila melanogaster.
The finding of a QuartOP in a mitochondrial genome quartet that supports a non-traditional grouping could either reflect a rare recombination event, selection pressures that led to convergent evolution in two lineages, or a chance event – if one looks at enough samples one will find some that (considered by themselves) appear significant. At present it is not possible to decide between these three possible explanations. Our analysis of mitochondrial genomes shows that in most instances the calculated probabilities (ML-mapping, Bayesian posterior probabilities, or bootstrap values) support the expected tree topologies, albeit with surprisingly strong support values. Rarely, unexpected groupings can be recovered and support for these probably erroneous groupings can be high. In most instances the ML-mapping approach accurately revealed the expected relationships between the mitochondrial genomes. This confirms the suitability of this approach in genome analyses.
Interphylum genome quartets
Here we focus on examples that illustrate the utility of the probability mapping approach. Focusing on the relationships between the cyanobacteria with other bacterial phyla we calculated several genome quartets that include the Synechocystis sp. genome and three members each of other bacterial phyla (see Table 2). In all cases that included both Bacillus subtilis as a representative of the low GC Gram positives, and Synechocystis sp., the majority of QuartOPs supported the topologies that grouped these two organisms together. The alternative topologies were significantly supported by some QuartOPs, but the number of strongly supporting QuartOPs was lower than for the B. subtilis – Synechocystis grouping. This also was true when one of the other two genomes was from a high GC Gram positive (genome quartets #53, #54, #55 and #68). Only when two low GC Gram positives were included in the same quartet, was the intra-phylum grouping of low GC Gram positives supported by many more QuartOPs than the grouping of B. subtilis with Synechocystis sp. (genome quartets #51 and #52).
<p>Table 2</p>
Summary of the genome quartets that include Synechocystis sp., Bacillus subtilis, and two bacterial genomes from other phyla.
#
Genome 1
Genome 2
Genome 3
Genome 4
((1,2),3,4)
((1,3),2,4)
((1,4),2,3)
Tot
0.9
0.99
Tot
0.9
0.99
Tot
0.9
0.99
9
Synechocystis
P. aeruginosa
D. radiodurans
B. subtilis
94
76
63
101
73
57
186
158
126
10
Synechocystis
P. aeruginosa
T. pallidum
B. subtilis
69
54
50
51
33
28
102
80
67
14
Synechocystis
R. sphaeroides
B. subtilis
E. coli
65
53
44
286
263
248
28
25
17
15
Synechocystis
R. sphaeroides
B. subtilis
D. radiodurans
95
72
60
201
173
149
73
60
47
16
Synechocystis
D. radiodurans
B. subtilis
T. pallidum
63
50
40
94
74
63
60
46
35
17
Synechocystis
B. subtilis
T. pallidum
E. coli
93
72
66
55
43
34
66
49
39
18
Synechocystis
D. radiodurans
T. pallidum
E. coli
86
68
54
65
49
38
75
54
47
19
Synechocystis
D. radiodurans
B. subtilis
E. coli
129
105
86
156
131
104
98
76
62
50
Synechocystis
B. subtilis
E. coli
M. loti
276
255
228
44
31
21
54
43
38
53
Synechocystis
B. subtilis
E. coli
M. leprae
125
104
82
101
84
65
119
19
66
54
Synechocystis
B. subtilis
E. coli
M. tuberculosis
141
114
97
101
82
69
128
101
89
55
Synechocystis
B. subtilis
M. loti
M. tuberculosis
189
164
139
92
74
59
80
58
44
64
Synechocystis
C. trachomatis
B. subtilis
T. pallidum
47
31
24
108
96
86
36
25
18
67
Synechocystis
C. trachomatis
B. subtilis
M. loti
64
48
32
116
104
84
72
51
44
68
Synechocystis
C. trachomatis
B. subtilis
M. tuberculosis
77
55
45
94
80
62
68
52
38
51
Synechocystis
B. subtilis
E. coli
S. aureus
33
19
15
361
349
15
7
5
52
Synechocystis