To score the similarity between a de novo motif and the most similar known motif, we first computed the normalized Euclidean distance between the de novo motif and all the known motifs in our collection using the optimal local alignment with at least five overlapping bases. We then selected the known motif with the lowest distance as the closest match. The similarity score was computed as the negative of the Z-score for this distance using the distribution of distances for all the motifs in the collection.
