Dissecting the binding mechanisms of transcription factors to DNA using a statistical thermodynamics framework.

At the heart of gene regulation are Transcription Factors (TFs), proteins which bind to DNA in a sequence specific manner and drive the activation or repression of genes. Statistical thermodynamics has shown to be a promising avenue to describe the binding mechanisms of TFs. Here, I present ChIPanalyser, an R/Bioconductor package that models and predicts binding of TFs to DNA using a statistical thermodynamic framework. First, I show that goodness of fit metrics are an important consideration in TF binding predictions as well as demonstrate ChIPanalyser’s high performance compared to other tools and frameworks. Then, I focused on investigating the binding mechanisms of three TFs that are known architectural proteins CTCF, BEAF-32 and su(Hw) in three Drosophila cell lines (BG3, Kc167 and S2). I demonstrate that architectural proteins show varying affinities towards DNA accessibility and that protein abundance plays a lesser role in their binding. While BEAF-32 binds in open chromatin, CTCF and su(Hw) showed increased binding is less accessible DNA. Furthermore, the model was able to recover binding preferences of three Hox TFs with respect to DNA accessibility. However, DNA accessibility showed some limitations to describe the full scope of TF binding affinities. I developed a genetic algorithm to investigate the binding affinity of the aforementioned TFs with respect to chromatin states. The improved model recovered chromatin state affinities and showed a more nuanced picture of TF binding. Finally, I examined the binding mechanisms of Su(H). The model was able to recover known binding mechanisms with respect to both chromatin state affinity and TF abundance. Overall, ChIPanalyser provides accurate TF binding predictions as well as insights into the mechanisms of TF binding.