Mapping Transcription Factor Networks By Comparing Tf Binding Locations To Tf Perturbation Responses

Abstract

26 Background: A transcription-factor (TF) network map indicates the direct, functional targets of 27 each TF --the genes it regulates by binding to their cis-regulatory DNA. Data on the genomic 28 binding locations of each TF and the transcriptional responses to perturbations of its activity, 29 such as overexpressing it, could support TF network mapping. Systematic data sets of both 30 types exist for yeast and for human K562 and HEK293 cells. 31 Results: In previous data, most TF binding sites appear to be non-functional, so one cannot 32 take the genes in whose promoters a TF binds as its direct, functional (DF) targets. Taking the 33 genes that are both bound by a TF and responsive to a perturbation of it as its DF targets 34 (intersection algorithm) is also not safe, as we show by deriving a new lower bound on the 35 expected false discovery rate of the intersection algorithm. When there are many non-functional 36 binding sites and many indirect targets, non-functional sites are expected to occur in the cis-37 regulatory DNA of indirect targets by chance. Dual threshold optimization, a new method for 38 setting significance thresholds on binding and response data, improves the intersection 39 algorithm, as does post-processing perturbation-response data with NetProphet 2.0. A 40 comprehensive new data set measuring the transcriptional response shortly after inducing 41 overexpression of a TF also helps, as does transposon calling cards, a new method for 42 identifying TF binding locations. 43 Conclusions: The combination of dual threshold optimization and NetProphet greatly expands 44 the high-confidence TF network map in both yeast and human. In yeast, measuring the 45 response shortly after inducing TF overexpression and measuring binding locations by using 46 transposon calling cards improve the network synergistically. 47 48

Recent advances in exploring transcriptional regulatory landscape of crops

“…However, it is important to note that these two evidence sources rarely converge on a common set of target genes. Despite being widely used as the gold standard, even the bound and differentially expressed genes may not be the validated functional targets (Kang et al, 2020).…”

“…This method improves the accuracy of identifying direct functional targets by combining data from TF binding sites and TF perturbation responses. The DTO method enhances the convergence of two data types by optimizing the significance thresholds for binding and responsive data (Kang et al, 2020). Morin and colleagues built upon two existing strategies (Tang et al, 2011;Wang et al, 2013) to create a framework to identify and rank TF-target interactions, and identified potential orthologous interactions between humans and mice.…”

Time-resolved genome-scale profiling reveals a causal expression network

“…Alternatively, genetic perturbations with expression or growth-rate as readouts can be used to group functionally similar genes and processes [8,9] . Gene expression profiling of deletion mutants (i.e., asymptotic readouts of a perturbation) can help identify co-regulated genes, though without dynamics there is limited potential for determining direct regulatory relationships because how information propagates from the deletion to each differentially expressed gene is not observed [10] .…”

FUN-PROSE: A Deep Learning Approach to Predict Condition-Specific Gene Expression in Fungi

“…Deciphering the complicated process of gene regulation was one of the key research objectives during the past several decades [8][9][10] . Due to the synchronized work of multiple systems controlling mRNA synthesis and decay 11,12 , the average steady state levels of individual mRNAs may vary from less than one copy per cell to several hundreds per cell across different environmental conditions 13 .…”