Table 2 Comparison of prediction performance for Nearest-Centroid classifiers built from unique VIP genes, top 50 p-value ranked genes, and 50 VIP genes. The classifier performance metrics, including accuracy (Acc), specificity (Spec), Sensitivity (Sens), and Matthew's correlation coefficient (MCC) were calculated based on averages of 50 repetitions of sample splitting, gene selection, and validation dataset prediction.
From: Very Important Pool (VIP) genes – an application for microarray-based molecular signatures
Unique VIP genes | 50 p-value ranked genes | 50 VIP genes | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Data set | Number of genes | Acc (%) | Spec (%) | Sens (%) | MCC | Acc (%) | Spec (%) | Sens (%) | MCC | Acc (%) | Spec (%) | Sens (%) | MCC |
Beer | 15 | 64.7 | 38.3 | 74.0 | 0.13 | 64.7 | 38.3 | 74.0 | 0.12 | 65.2 | 35.4 | 75.6 | 0.11 |
Bhattacharjee | 17 | 58.7 | 57.8 | 59.6 | 0.18 | 58.0 | 59.2 | 56.8 | 0.16 | 58.6 | 59.4 | 57.8 | 0.18 |
Chen | 14 | 96.5 | 99.9 | 93.6 | 0.93 | 95.3 | 100.0 | 91.2 | 0.91 | 95.8 | 100.0 | 92.1 | 0.92 |
Pomeroy | 20 | 60.8 | 54.0 | 64.2 | 0.19 | 60.8 | 51.7 | 65.3 | 0.18 | 62.4 | 55.7 | 65.8 | 0.22 |
Rosenwald | 18 | 55.5 | 58.1 | 53.6 | 0.12 | 56.8 | 63.2 | 52.2 | 0.15 | 57.4 | 62.3 | 53.8 | 0.16 |
Shipp | 18 | 51.6 | 50.8 | 52.5 | 0.03 | 47.9 | 51.8 | 43.0 | -0.05 | 49.0 | 47.4 | 51.0 | -0.02 |
Singh | 15 | 94.3 | 98.3 | 91.7 | 0.89 | 98.1 | 100.0 | 96.9 | 0.96 | 97.8 | 100.0 | 96.4 | 0.96 |
Yeoh | 22 | 74.6 | 37.8 | 80.2 | 0.15 | 78.2 | 31.0 | 85.4 | 0.15 | 80.2 | 35.0 | 87.0 | 0.21 |
van't Veer | 20 | 64.8 | 64.8 | 64.9 | 0.30 | 65.2 | 61.5 | 68.6 | 0.31 | 66.9 | 66.1 | 67.6 | 0.34 |