Branch | Free-ratio | 2-ratio |
---|
 | d
N
/d
S
| Bootstrap | d
N
/d
S
| 2ΔL |
P
|
---|
A | 0.089 | 74 | 3.883 | 2.942 | 0.0862 |
B | 0.104 | 100 | 0.150 | 0.713 | 0.3985 |
C | 0.105 | 100 | 0.097 | 0.004 | 0.9522 |
D | 0.097 | 99 | 0.108 | 0.055 | 0.8140 |
E | 0.076 | 100 | 0.063 | 0.293 | 0.5883 |
F | < 0.001 | 97 | < 0.001 | 2.031 | 0.1541 |
G | 0.212 | 79 | 0.229 | 0.642 | 0.4229 |
H | 0.047 | 66 | 0.204 | 0.238 | 0.6258 |
I | 1.454 | 33 | 0.219 | 0.531 | 0.4663 |
J |
> 999
| 79 |
> 999
| 7.296 | 0.0069# |
K | < 0.001 | 84 | < 0.001 | 2.161 | 0.1416 |
L | 0.041 | 100 | 0.038 | 2.794 | 0.0946 |
M | < 0.001 | 98 | < 0.001 | 0.249 | 0.6180 |
N |
3.493
| 87 |
2.572
| 7.653 | 0.0057*,# |
O | 0.056 | 100 | 0.044 | 1.831 | 0.1760 |
P | 0.061 | 66 | 0.036 | 2.259 | 0.1329 |
Q | 0.009 | 100 | 0.009 | 2.020 | 0.1552 |
- Branch names are as depicted in Fig. 4B. For the first comparison, d
N
/d
S
rate ratios were inferred for individual branches with the free-ratio model, in which all branches were allowed to vary; the optimal model had a likelihood of ln L = -2871.29; the significance of individual branches having a d
N
/d
S
rate ratios above 1 or below 1 was assessed with non-parametric bootstrapping using 100 replicates; d
N
/d
S
rate ratios larger than 1 and with bootstrap support of more than 50 are given in bold. For the second comparison, d
N
/d
S
rate ratios were repeatedly inferred with the 2-ratio model, in which only the branch of interest was allowed to differ from the remaining branches; the significance of the individual branches to be different from the remaining branches was assessed via a likelihood ratio test comparison to the null model, in which all branches of the tree were assumed to have identical d
N
/d
S
rate ratios; the null model had a likelihood score of ln L = -2895.56; the probability P was calculated from twice the likelihood difference 2ΔL between null model and tested model; significance is indicated by * and ** for α = 0.1 and 0.05 according to the sequential Bonferroni procedure, respectively, and by # and ## for α = 0.1 and 0.05 according to the false discovery rate, respectively; bold d
N
/d
S
rate ratios indicate those that are significantly larger than 1 according to either method.