Figure 7

The influence of the number of suboptimal structures on accuracy. The performance of the algorithm as a function of the number of suboptimal predictions used as input for our eight datasets. For most datasets, the results improve substantially for the first 50–150 predictions. A further increase in the number of suboptimal structures does not yield much better accuracy. Hence, a value of 150 predictions was chosen as the default for the RNAspa program.