Transcription factor recognition models with modified nucleobases

点击翻译按钮获取中文摘要

Transcription factors drive mammalian gene expression programs, initiating transcription by binding DNA in specific sequence and epigenomic contexts. Sequence specificity of transcription factors has been studied extensively for decades. The study of epigenomic contexts became possible only later with large-scale data on post-translational histone modifications and DNA methylation. In methylated DNA, cytosine (C) undergoes conversion to a modified form, 5-methylcytosine (5mC). DNA methylation can influence transcription through non-sequence-specific 5mC-binding proteins which can silence genes. Nonetheless, some transcription factors preferentially bind specific sequences containing 5mC rather than unmodified C, or vice versa. The cell maintains stable DNA methylation states over time and even across cell divisions. This means methylation can exhibit a more sustained impact on transcription factor binding than other, more transient, epigenomic marks. In addition to 5mC, recent studies reveal the biological importance of three other cytosine modifications, 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC). Other studies show that these modifications also affect transcription factor recognition of DNA just as 5mC does. There are no tools, however, to predict transcription factor affinity for binding sites that consider the presence or absence of modified nucleobases. This prevents a full understanding of the impact of cytosine modifications on gene regulation.***To remove this barrier to the understanding of gene regulation, we will develop and validate computational models and methods to determine transcription factor binding affinities in the context of epigenetic nucleobase modifications. We will accomplish this through the following specific aims: ***Aim 1. We will create an algorithm to determine likely DNA modification state using multiple evidence sources.***Aim 2. We will construct a new mathematical model of transcription factor binding preference for sequences of modified and unmodified nucleotides.****Aim 3. We will develop a hybrid approach for motif discovery that expands in vitro transcription factor binding models to the broader alphabet using in vivo chromatin immunoprecipitation-sequencing (ChIP-seq) data.***The applicant, Dr. Michael Hoffman, has formal training in machine learning, bioinformatics, and biochemistry. He has published multiple papers on transcription factor binding site prediction and epigenetic state interpretation. He has managed the development of sustainable software used by research users in multiple countries daily.***This research program will lead to a better understanding of the relationship between nucleobase modifications and gene regulation. It will also deliver software, genome annotations, and model files to enable the genomics community to use our methods and results.**