Fig. 3.

Identification of features associated with AR enhancers. a Features from all publicly available transcription factor and histone mark ChIPseq datasets in LNCaP cells (GEO accession and the citations provided in supplementary data) were uniformly processed and binding energy was calculated. The bar graph shows the features sorted by their binding energy at inducible enhancers. The zoomed section (top right) shows the top 5 features that are predictive of inducible enhancers. b LNCaP ARBS predicted by the machine learning classifier as either inactive or inducible enhancers were validated by luciferase assay (4 biological replicates±SEM). c Receiver operating characteristic curve of AR and H3K27ac ChIPseq to accurately identify inducible enhancers