Figure 4

Impact of size of training set on the classification of gap-free genomic alignment blocks. For each proportion of the original training set used (x axis) 100 randomly selected training sets were generated and SVM parameters optimized independently. Sensitivity in training (1% false positive threshold) and sensitivity and false positive rates at the same threshold with validation data are shown. Blue = sensitivity long CSTs - training, Red = sensitivity long CSTs - test, Dark blue = False positive rate long CSTs - test, Yellow = sensitivity short CSTs- training, Green = sensitivity short CSTs - test, Orange = false positive rate short CSTs - test.