Fig. 3

TBA, a transcription factor binding analysis. a Schematic workflow of TBA. Binding sites for a transcription factor (green boxes) are mixed with random GC-matched background sequences (gray boxes). Motifs from the JASPAR library are merged to create a non-redundant motif library. Motif scores are calculated for all sequences at all binding sites and GC-matched background and then used to train a TBA model. Model weights from the trained model indicate whether a motif is positively or negatively correlated with the occupancy of a transcription factor. The performance of the full model and a perturbed model with one motif removed are compared to identify motifs that are important to the model. The intersection of important motifs that affect model performance and the model weights learned by the classifier can be used to infer the binding partners of a transcription factor. b, c Distribution of variance inflation factor for each motif in the TBA merged motif library and JASPAR motif library for experiments performed in b Vehicle and c Kdo2 lipid A treated thioglycollate elicited macrophages