Table 4.
Quantitative characteristics of DT models and ILP theories
| |
DT (# nodes per model) |
ILP (# rules per theory) |
||
|---|---|---|---|---|
| Avg (min-max) | % total | Avg (min-max) | % total | |
| Model coverage (%) |
58 (32–67) |
- |
83 (77–88) |
- |
| Model size |
11 (6–15) |
- |
33 (16–40) |
- |
|
Drug descriptors |
|
|
|
|
| Categories |
4 (1–7) |
34 |
6 (2–13) |
19 |
| Targets |
3 (0–5) |
26 |
30 (23–39) |
90 |
| Clusters |
4 (1–9) |
40 |
9 (4–14) |
27 |
|
Target descriptors |
|
|
|
|
| GO terms |
NA |
NA |
24 (16–31) |
73 |
| Domains |
NA |
NA |
1 (0–2) |
1 |
| Interactions |
NA |
NA |
8 (2–16) |
24 |
| Pathways |
NA |
NA |
4 (1–8) |
12 |
| GO relationships | NA | NA | 6 (3–9) | 19 |
Model coverage is the percentage of positive examples covered, averaged over the 26 DT models and 26 ILP theories. Avg: average. Model size corresponds to the average number of nodes in a DT model or of rules in a ILP theory. Occurrence of each type of descriptor is estimated by counting the number of nodes (rules respectively) involving them (NA: not applicable).