Genes selected | Accuracy (%)a
| MCC^
|
---|
SVM | RF | SVM | RF |
---|
All genesb
| 95.3 | 76.9 | 0.91 | 0.00 |
GSVA feature enrichment | 98.5 | 76.9 | 0.87 | 0.00 |
sRAP | 100 | 76.9 | 1.00 | 0.00 |
SVM-RFE | 100 | 100 | 1.00 | 1.00 |
RF-based Positive MDA | 100 | 76.9 | 1.00 | 0.00 |
T-test | 100 | 97.0 | 1.00 | 0.91 |
- The accuracy of the SVM predictors were obtained from LOO cross validation. SVM and RF classifiers were constructed with each set of data listed in Table 2
-
aAll percentages are rounded off to three significant figures
-
bTranscripts with a total expression of zero and/or having more than six samples with expression levels less than one were excluded
-
^Matthews correlation coefficient (MCC) rounded to 2 decimal places