In peak-motifs, we used the default background model: build from the input peak sequences with a Markov order adapted according to the sum of sequence sizes (option -auto)
In some cases, this returns highly significant motifs that are not specific for the particular peak set analysed, but are generally over-represented in the genome, or in regulatory sequences.
A more suited approach to detect motifs specific for the considered peakset would be to build a background model from a "neutra" peakset, i.e. the test sequences provided in IBIS leaderboard or final board.
In peak-motifs, we used the default background model: build from the input peak sequences with a Markov order adapted according to the sum of sequence sizes (option
-auto)In some cases, this returns highly significant motifs that are not specific for the particular peak set analysed, but are generally over-represented in the genome, or in regulatory sequences.
A more suited approach to detect motifs specific for the considered peakset would be to build a background model from a "neutra" peakset, i.e. the test sequences provided in IBIS leaderboard or final board.