Multiple Motif Detection in DragoNN Simulated Data

[GoktugGuvercin]

Hello;

While I was working on transcription factor binding sites and motif detection, I noticed your DragoNN toolkit and Github profile. It is very informative and useful. I aim to develop a deep learning model for multiple motif recognition. At this point, I intend to use your simulation data accessible via the following link: https://github.com/kundajelab/dragonn/blob/master/paper_supplement/simulation_data/GC_fraction0.4max_num_motifs3min_num_motifs0motif_names%5B'CTCF_known1'%2C%20'ZNF143_known2'%2C%20'SIX5_known1'%5Dnum_seqs20000seq_length500.npz

As far as I understand, the sequences in this dataset consists of 500 nucleotides and total percentage of guanine-cytosine in the sequences is approximately 0.4 However, I am confused at the number of motifs in the sequences. Max and minimum number of motifs are set to 3 and 0 respectively.

What does it exactly mean ?

Can 3 instances of each motif exist in a positive sequence ? In other words, the max number of motifs is for each motif or sum of all three motifs ? In first case, up to 3 instances of each motif can exist in a positive sequence. In second case, only 1 instance of each motif can be accommodated in a positive sequence.