Detecting Differential Transcription Factor Activity from ATAC-Seq Data

Abstract

Transcription factors are managers of the cellular factory, and key components to many diseases. Many non-coding single nucleotide polymorphisms affect transcription factors, either by directly altering the protein or its functional activity at individual binding sites. Here we first briefly summarize high-throughput approaches to studying transcription factor activity. We then demonstrate, using published chromatin accessibility data (specifically ATAC-seq), that the genome-wide profile of TF recognition motifs relative to regions of open chromatin can determine the key transcription factor altered by a perturbation. Our method of determining which TFs are altered by a perturbation is simple, is quick to implement, and can be used when biological samples are limited. In the future, we envision that this method could be applied to determine which TFs show altered activity in response to a wide variety of drugs and diseases.

A stable mode of bookmarking by TBP recruits RNA polymerase II to mitotic chromosomes

“…For all transcription factor motif models in mm10 (each dot), the change in MD-score in Async vs Mitosis (left), untreated vs IAA in Async cells (middle), and untreated vs IAA in Mitotic cells (right), are plotted relative to the number of motifs within 1.5 kb of any ATAC-seq peak center (x-axis). MD-scores are defined as the enrichment of a TF sequence motif within a small radios (150 bp) of ATAC-seq peaks relative to a larger local window (1500 bp) ( Tripodi et al, 2018 ). Significantly different MD-scores are highlighted in red and purple (p-value<1×10 −6 ).…”

From reads to insight: a hitchhiker’s guide to ATAC-seq data analysis

“…Open chromatin can be detected by piling up short fragments from NFRs or using a shift-extend approach, which tries to count the cutting events smoothed by the extension size ( Fig. 3b, right box) [61,62]. This approach is more generic, as it can be applied to almost all ChIP-seq peak callers and is not affected by the fragment size of data.…”

“…MEME-CentriMo [121] further identifies motifs enriched near peak centers. DAStk [62] generates a MD score (motif displacement score) [122]. This is achieved by calculating the ratio of motif occurrence within a small window (150 bp) to a large radius (1500 bp) from each peak center.…”

ATAC-seq signal processing and recurrent neural networks can identify RNA polymerase activity

“…The different types of quality control warnings could also be used to weight the different machine learning classification approaches accordingly in an ensemble setting. Another interesting next step would be to explore how tools that employ ATAC-seq data for differential analysis of TF activity such as DAStk (Tripodi et al, 2018b) can benefit from using one of these classifiers as a pre-filter, by using only those peaks considered as "overlapping active Pol II recruitment", instead of all peaks from each sample. Alternatively, DAStk could weight peaks based on their classification outcome, giving those predicted to overlap RNA polymerase larger weights.…”