Table 6.
SVM Analysis of Expression Data
| Training set | Number of ESTs | 2500 |
| Number of positive | 157 | |
| Number of negative | 2341 | |
| Number of positive support vectors | 131 | |
| Number of negative support vectors | 139 | |
| Testing set | Number of ESTs | 1387 |
| True positive | 44 | |
| False positive | 10 | |
| True negative | 1286 | |
| False negative | 47 | |
| Prediction set | Number of ESTs | 7562 |
| Number of positive predicted | 110a | |
| Number of negative predicted | 7452 |
Parameter of the program, C = 10; Gaussian radial basis function kernel with δ = 0.01. (SVM) Support vector machine.
a
80 of the 110 predicted positive ESTs were sequenced and they belong to 56 unique cDNA clusters. The remaining 30 cDNAs were not sequenced. Details in Supplementary Figure 6. Known muscle ESTs were used as positives and selected known nonmuscle ESTs were used as negatives.