Table 1. A comparison of the algorithms on simulated mutation data with varying passenger mutation probability 
.
Avg. distance  from planted pathways |
|||
|
Multi-Dendrix | Iter-RME | Iter-Dendrix |
| 0.0 |
|
0.01
0.12
|
|
| 0.0001 |
0.02
0.18
|
0.01
0.16
|
0.30
0.86
|
| 0.0005 |
0.04
0.23
|
0.10 0.40 |
0.35 0.89 |
| 0.001 |
0.10
0.35
|
0.32
0.60
|
0.44
1.01
|
| 0.005 |
0.44
0.71
|
– | 0.75 1.07 |
| 0.01 |
1.03
1.00
|
– | 1.20 1.15 |
| 0.015 |
1.68
1.16
|
– | 1.78 1.26 |
| 0.02 |
2.17
1.24
|
– | 2.21 1.29 |
Italicized rows correspond to values of 
approximated from real cancer datasets. Each entry is mean (
) and standard deviation (
) (across 1000 simulations) of the distance 
between the planted set of pathways 
and the collections 
found by each algorithm. The minimum distance 
indicates an algorithm found the planted pathways exactly, while the maximum distance 
indicates that an algorithm did not find any of the genes in the planted pathways. Bold text indicates the top performing algorithm for each value of 
. Multi-Dendrix is the top performer for all values of 
except the smallest 
. The differences between Multi-Dendrix and both Iter-Dendrix and Iter-RME are statistically significant (
) for 
. For 
, Iter-RME did not complete after 24 hours of runtime.