Characterisation of human transcription factors binding mechanisms in human cell lines using a statistical thermodynamic framework

As key genome regulatory elements, transcription factors (TFs) bind to the DNA to control gene expression. High-throughput technologies such as ChIP-seq can experimentally determine TF binding; however, the raw data alone cannot fully answer questions such as whether a TF found in open chromatin is the one responsible for the relaxed state (thus displaying pioneering properties), or it came along and bound an already accessible site. In this study we wanted to investigate the interplay between TFs and chromatin accessibility. I used ChIPanalyser, a Bioconductor tool, to model and predict ChIP-seq-like profiles in human cell lines, based on publicly available ChIP-seq and DNA accessibility datasets. I estimated the binding parameters of twenty TFs in IMR90 and HepG2 cell lines and used their profiles to evaluate their preference for (or lack of) open and dense chromatin. Our analysis supports that there are a number of TFs (e.g. CTCF) that display the same properties whether we consider the DNA to be accessible or not, highlighting that some TFs are insensitive to chromatin accessibility, and this holds true for regions with weaker binding sites as well as strong. Our results also suggest there are subsets of TFs that have a preference for open chromatin (e.g. MAZ). Out of the 20 TFs analysed, 4 displayed pioneering functions when looking only at the strong bound regions and 3 at both strong and medium-bound regions. These results are true when using different accessibility measure methods, like ATAC-seq, DNase-seq, MNase-seq and NOMe-seq, hence the results are not method-specific. In short, ChIPanalyser can be used to model and predict ChIP-seq data and learn new biological insights, to predict TF binding events between cell lines, or for a screening process to understand a TF’s behaviour.