DMFS pipeline recovery of previously identified motifs.

Here we list motifs identified by Tillo and Hughes [48] and Lee et al.[49] and the number of times these motifs were identified by the DMFS pipeline. Structure related features are omitted, as are transcription binding start sites and features with zero weights. We ran the DMFS pipeline 40 times, with random data partitioning, and counted the number of times each previously identified motif occurred. According to Tillo and Hughes the most discriminative motif is the 4-mer AAAA/TTTT, which emerged in almost every run.