Figure 5

Figure a) shows the average weight (over 10 runs) of the weights per position (one weight for two positions) and d) the averaged weights per oligomer length (uniform position weighting). Figures b) displays the position and oligomer length combinations that were found to be significantly used (40 bootstrap runs). Figure c) shows the likelihood for rejecting ${\mathscr{H}}_0$. In all runs we used 5, 000 training examples.