Predicting variation of DNA shape preferences in protein-DNA interaction in cancer cells with a new biophysical model

TL;DR

This paper introduces a new biophysical model that incorporates DNA shape features to better understand transcription factor-DNA interactions, especially variations in cancer cells, aiding in mutation effect prediction.

Contribution

A novel biophysical model based on neighbor dinucleotide dependency that integrates DNA shape features for analyzing transcription factor binding in cancer cells.

Findings

DNA shape preferences vary at FOXA1 sites after steroid treatment in MCF7 cells.

The model provides biophysical interpretation of DNA shape preferences.

Potential to improve prediction of mutation effects in cancer.

Abstract

DNA shape readout is an important mechanism of target site recognition by transcription factors, in addition to the sequence readout. Several models of transcription factor-DNA binding which consider DNA shape have been developed in recent years. We present a new biophysical model of protein-DNA interaction by considering the DNA shape features, which is based on a neighbour dinucleotide dependency model BayesPI2. The parameters of the new model are restricted to a subspace spanned by the 2-mer DNA shape features, which allowing a biophysical interpretation of the new parameters as position-dependent preferences towards certain values of the features. Using the new model, we explore the variation of DNA shape preferences in several transcription factors across cancer cell lines and cellular conditions. We find evidence of DNA shape variations at FOXA1 binding sites in MCF7 cells after treatment with steroids. The new model is useful for elucidating finer details of transcription factor-DNA interaction. It may be used to improve the prediction of cancer mutation effects in the future.