Figure 2.

Structure-derived allele preference metrics show capacity to distinguish between differentially bound allele pairs, but the performance is highly heterogenous across TFs. Analyses represent ROC curves, with AUC representing the classification performance. (a) Generalized metric performance in distinguishing between pbSNP containing pairs and nondifferentially bound sequences (non-pbSNPs). Analysis represents allele pairs pooled from all six TFs: SPIB, ETV4, ELK3, SF-1, PAX5, and MEIS2. (b) TF-wise classification performance. Qualitative binding preference groups were annotated in line with the SNP-SELEX paper classification. Structural metric directions were adjusted to match the directional convention in the analyses.