Table 2.
Performance of SVM-Radial and PAM Models in Independent Dataset Class Prediction Analysis
| SVM-radial | |||||
| Features | Accuracy | Sensitivity | Specificity | Precision | AUROC |
| Top5 | 54.44 (0.06) | 66.05 (0.12) | 37.24 (0.13) | 60.87 (0.04) | 53.98 (0.05) |
| Top25 | 60.56 (0.04) | 71.4 (0.06) | 44.48 (0.15) | 66.12 (0.05) | 60.60 (0.06) |
| Top100 | 67.08 (0.04) | 75.35 (0.05) | 54.83 (0.1) | 71.50 (0.04) | 73.34 (0.05) |
| Top400 | 67.64 (0.04) | 76.28 (0.05) | 54.83 (0.09) | 71.69 (0.04) | 75.26 (0.04) |
| Top1600 | 71.39 (0.03) | 75.81 (0.05) | 64.83 (0.06) | 76.27 (0.03) | 75.27 (0.02) |
| All genes | 70.83 (0.03) | 66.98 (0.08) | 76.55 (0.06) | 81.22 (0.03) | 82.57 (0.02) |
| PAM | |||||
| Top5 | 61.53 (0.05) | 69.07 (0.09) | 50.35 (0.10) | 67.44 (0.04) | 63.56 (0.04) |
| Top25 | 65.56 (0.04) | 64.42 (0.07) | 67.24 (0.04) | 74.36 (0.03) | 70.76 (0.03) |
| Top100 | 64.45 (0.03) | 63.26 (0.03) | 66.21 (0.07) | 73.67 (0.04) | 74.91 (0.02) |
| Top400 | 67.09 (0.01) | 63.72 (0.02) | 72.07 (0.03) | 77.20 (0.02) | 75.40 (0.02) |
| Top1600 | 67.22 (0.02) | 63.26 (0.03) | 73.10 (0.04) | 77.78 (0.02) | 74.89 (0.01) |
| All genes | 67.09 (0.02) | 62.56 (0.03) | 73.79 (0.02) | 77.98 (0.02) | 73.92 (0.01) |
Notes: Support vector machine (SVM)-radial and prediction analysis of microarrays (PAM) models were developed using different number of top differentially expressed genes (DEGs). These models were evaluated for their performance based on their ability to predict samples of GSE27383 as independent dataset. The models were compared using RMA followed by Tukey’s post hoc test with Greenhouse-Geisser and Huynh-Feldt corrections for each parameter separately. SVM-radial models performed better with higher number of feature DEGs. The values mentioned in the tables are in percentage and the standard deviation for the ten iterations in bracket.