Table 2. Performance metrics of classifiers on the lung cancer test set (AC and SCC subtype classification).
| The data used (Total # of samples) | N# of Genes | Error (%) | GBS (0) | BCM (1) | AUPR (1) | |
| Ben-Hamo's study | GSE10245, GSE18842, GSE31799 (151, 81AC, 70SCC) | 1 | 15.3 | NA | NA | NA |
| TGDR | GSE10245, GSE18842, GSE2109, GSE31908 (175, 100AC, 75SCC) | 20 | 16 | 0.1153 | 0.8325 | 0.9416 |
| A. Radiz on 3-gene signature selected by AC and SCC subtype classification | ||||||
| Radviz alone | GSE10245, GSE18842, GSE2109 (only stage I &II, 145, 71AC, 74SCC) | 3 | 16.67 | – | – | – |
| Radviz +TGDR | GSE10245, GSE18842, GSE2109 (145) | 3 | 14.67 | 0.2360 | 0.5144 | 0.8917 |
| Radviz+naïve Bayes | GSE10245, GSE18842, GSE2109 (145) | 3 | 13.33 | 0.1260 | 0.8447 | 0.8908 |
| Radviz+SVM | GSE10245, GSE18842, GSE2109 (145) | 3 | 13.33 | 0.1208 | 0.6974 | 0.8978 |
| B. Radiz on 8-gene signature selected by subtype & stage classification | ||||||
| Radviz alone | GSE10245, GSE18842, GSE2109 (145) | 8 | 14 | – | – | – |
| Radviz +TGDR | GSE10245, GSE18842, GSE2109 (145) | 8 | 13.33 | 0.1061 | 0.8271 | 0.8935 |
| Radviz+naïve Bayes | GSE10245, GSE18842, GSE2109 (145) | 8 | 12.67 | 0.1191 | 0.8719 | 0.9067 |
| Radviz+SVM | GSE10245, GSE18842, GSE2109 (145) | 8 | 14 | 0.1029 | 0.7983 | 0.9211 |
NA: not available. –: not computable because no posterior probabilities were provided.