Table 3.
Evaluation of the performance of our system using SVM-based PRE classifier vs BART-based PRE classifier
| Method | EpiPredictor components | Schwartz et al.a | Tolhuis et al.b | Schuettgurber et al.c | Intersectiond |
|---|---|---|---|---|---|
| SVM | (a, b) | 22.73% | 9.78% | 19.53% | 23.68% |
| (a, b, c) | 26.14% | 10.22% | 25.12% | 23.68% | |
| BART | (a, d) | 21.59% | 8.44% | 19.07% | 21.05% |
| (a, d, c) | 22.73% | 9.33% | 22.79% | 21.05% |
(a): Motif Analyzer; (b): SVM-based Classifier; (c): GC Analyzer; (d): BART-based Classifier.
aOverlap between the top 243 predicted genes with the genes predicted by Schwartz et al. (4).
bOverlap between the top 243 predicted genes with the genes predicted by Tolhuis et al. (15).
cOverlap between the top 243 predicted genes with the genes predicted by Schuettengruber et al. (14).
dOverlap between the top 243 predicted genes with the genes intersected by Schwartz et al., Tolhuis et al. and Schuettengruber et al. The results of the SVM-based classifier are highlighted in bold.