Skip to main content
. 2008 Jul 1;24(13):i96–i104. doi: 10.1093/bioinformatics/btn146

Table 1.

Performance comparison of sampling method and guided halving algorithm in the case of one outgroup

Halving analysis
 Sampling method
 Guided halving
R−T 2n dt,x⊕x dx,r dmin da,r ΔA da,a* Time dx,r da,r ΔA Inline graphic Time
AG-CG 538 186 204 196 180 −16 156 37 153 153 0 120 2.3
AG-SC 1012 119 237 229 208 −21 53 158 184 183 −1 32 5.3
KL-CG 546 186 210 203 184 −19 154 50 160 160 0 120 3.5
KL-SC 1026 122 241 232 216 −16 51 140 197 197 0 39 6.1
KW-CG 542 188 247 238 230 −8 167 26 216 215 −1 142 3.3
KW-SC 994 121 364 355 350 −5 70 72 325 323 −2 41 5.1
A*-CG 600 199 183 169 129 −40 129 81 84 84 0 84 1.5
A*-SC 1062 124 79 70 37 −33 37 114 5 5 0 5 0.3
AG-V 576 61 157 151 149 −2 54 12 148 148 0 51 0.9
KL-V 584 62 167 160 158 −2 53 12 157 157 0 51 0.9
KW-V 582 62 224 218 215 −3 52 13 212 212 0 51 1.0
A*-V 600 62 57 49 39 −10 39 14 29 29 0 29 0.2

Sample size 2000 for the sampling method. R−T represents the outgroup and doubling descendant. n is the number of genes available in that pair of genomes, with two copies in T. dt,x⊕x=d(T,X′⊕X′′) is the doubling distance, constant over all analyses. Inline graphic represents the average, over all samples, of the distance estimate between the ancestor, just before doubling, and the outgroup, and the adjacent entry dmin=minsampled(X,R) is the minimum found. ΔA is the improvement over d(T,X′⊕X′′)+d(X,R) due to local searching, allowing A to be found outside the set of halving solutions. da,a*=d(A, A*) is the distance between the inferred ancestor and the ‘ground truth’. Time is measured in minutes, for 2000 samples of unrestricted halving or for one GGH run.