Figure 1

Overview of PIPES. (a, b) Starting with raw PBM data for a TF represented as fluorescence intensities to each individual probes, we first infer binding probabilities to individual short k-mers, and then a given sequence can be scored by such inferred binding probabilities. (c, d) To predict in vivo TF binding, we take as input the tissue-specific DNase I hypersensitivity data (tag counts) and convert them to probabilities that represent chromatin accessibility for each position in the genome at each tissue/cell types. (e) The PBM data, DNase data and other types of data including sequence conservation are combined using an integrative model. See Methods for details.