Organizing Principles of Regulatory Elements and TF Binding Models

Mammalian genomes contain millions of putative regulatory sequences, which are delineated by the binding of multiple transcription factors (TFs). A major question in regulatory element biology is how TFs collaborate to encode cell-specific activity of cis-regulatory DNA. Genomics approaches have given rise to various theories about the organizing principles of individual regulatory elements. The enhanceosome model posits that an individual regulatory element is defined by a combination of TFs that adhere to strict spacing and orientation constraints (SAOCs). In contrast, the billboard model suggests that regulatory DNA has (at best) soft SOACs, and individual TF binding sites are subject to a high rate of 'turnover' during evolution. My thesis work aims to explore two major issues around this topic: First, while various lines of genomic evidence have been used to compel each of these models, a systematic evaluation of how well each model explains the global sequence and biophysical properties of regulatory elements has been lacking. Therefore, I evaluated whether four different regulatory element models incorporating different degrees of SAOC among transcription factor binding sites were broadly consistent with five different global properties of regulatory sequence (length, uniqueness, frequency, turnover, and role of master regulators). Based on my analysis, I conclude that models of transcription factor binding with or without spacing and orientation constraints predict that regulatory sequences should be fundamentally short, unique, and turn over rapidly. Additionally, my analysis suggested that the existence of master regulators may be, in part, a consequence of evolutionary pressure to limit the complexity and increase evolvability of regulatory sites. Second, in order to thoroughly explore multiple TF binding and the regulatory code, I assisted in the identification of binding motifs for the remaining TFs that lack motifs. I systematically compared traditional position weight matrix models of TF specificity to more complex k-mer based approaches like gapped-kmer support vector machine (gkmSVM) both within and across Chip-Seq and HT-SELEX datasets. I discovered that gkmSVM outperforms PWMs in predicting HT-SELEX and Chip-Seq data for almost all TFs tested, suggesting that TF binding may be influenced by flanking sequences and dinucleotide interdependency that are not accounted for by a PWM model. Additionally, I find that cross-comparing models across datasets is critical for identifying systematic biases in these datasets and flagging uninformative experiments. Overall, my research will contribute greatly to our understanding of the regulatory element code and will aid in the understanding of multiple TF binding and individual TF binding.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.