Fig. 3

Sequence logos for corresponding AMs and FMs show overall similarity with minor differences. Sequence logos are shown for three representative SRFs: TRA2A, a splicing activator, and splicing repressors HNRNPA1 and HNRNPA2B1. The first column shows the FM motif and the second shows AM logos for 2 replicate runs. The third column shows differences between each AM and the corresponding FM: each base’s height represents its pointwise mutual information, with those above the bar enriched in the AM and those below enriched in the FM. The pattern of differences is often but not always related between replicates, as shown by the examples above. The accuracy a logistic model can achieve at distinguishing AM from FM motifs is shown above each difference plot. The fourth column shows the overlap between FM and AM binding sites for each factor, labeled with the fraction of genomic positions for each category. The total fraction of genomic positions for each model is shown next to the model’s name in square brackets. For example, FMs and AMs for TRA2A bind 0.16% of genomic sites, i.e., one site every 625 nt