Figure 2
From: Uncovering uncharacterized binding of transcription factors from ATAC-seq footprinting data
Iterative motif discovery from complex sequence mixtures. (a) Simulated de novo motif discovery by MEME, each dot indicates the closest MEME created motif compared to the respective original motif; As indicated by black circles, there is a correlation of binding site count (y axis) and de novo motif similarity distance (x axis) to the original motif. (b) Scheme of the iterative motif prediction; Identified FP sequences are extracted to build the motif discovery pool and used to create de novo motifs (red box). Created motifs are filtered for significance and all corresponding sequences used to generate significant motifs are removed from the sequence discovery pool. This process is run iteratively until no further significant motifs are predicted. The figure was created with BioRender.com. (c) Number (y axis) of generated motifs per iteration (x axis), split by significance (orange and blue) i.e. motifs accepted by internal dynamic e-value filter. (d) Heatmap showing the distance of de novo motifs detected per iteration (columns) to the most similar database motif (row) for significant motifs from a simulated de novo run. To the left, the number of sites from the corresponding ChIP-seq run per factor is shown as a bar plot. Grey color indicates no significant motifs were available within one iteration.
