Skip to main content
. 2022 May 23;40(10):1520–1527. doi: 10.1038/s41587-022-01307-0

Extended Data Fig. 9. Comparison of GR binding models learned using different algorithms.

Extended Data Fig. 9

Top: Binding models inferred by peak-based methods (MEME-ChIP and HOMER) and peak-free methods (ProBound and NoPeak) from the GR ChIP-seq data published in Starick et al. (2015). For MEME-ChIP, the reverse-complement symmetry setting was activated. Bottom: Comparison of ChIP-based and SELEX-based binding models for GR, displayed as in Fig. 5a. Because the binding models generated by MEME-ChIP and HOMER contain base probabilities p, the negative logarithm of these values were compared to the ΔΔG/RT values from the SELEX model. None of the binding models found by NoPeak matched the GR consensus sequence.