Figure 2

Feature importance measure for sequence features. The discriminatory power of each feature was determined by calculating the importance value, with larger values indicating more relevant properties. The importance distribution is shown for each sequence feature as a box plot in which the middle bar is the median, the outer edges are the 10 and 90 percentiles and the edges of the box are the 25 and 75 percentiles. Outliers are shown as circles. Neg_1 (all experimental samples and inferred negative samples) and Neg_2 (all experimental samples and artificial negative samples from miRanda) were analyzed separately.