Identification of transcription factors dictating blood cell development using a bidirectional transcription network-based computational framework

Abstract

Advanced computational methods exploit gene expression and epigenetic datasets to predict gene regulatory networks controlled by transcription factors (TFs). These methods have identified cell fate determining TFs but require large amounts of reference data and experimental expertise. Here, we present an easy to use network-based computational framework that exploits enhancers defined by bidirectional transcription, using as sole input CAGE sequencing data to correctly predict TFs key to various human cell types. Next, we applied this Analysis Algorithm for Networks Specified by Enhancers based on CAGE (ANANSE-CAGE) to predict TFs driving red and white blood cell development, and THP-1 leukemia cell immortalization. Further, we predicted TFs that are differentially important to either cell line- or primary- associated MLL-AF9-driven gene programs, and in primary MLL-AF9 acute leukemia. Our approach identified experimentally validated as well as thus far unexplored TFs in these processes. ANANSE-CAGE will be useful to identify transcription factors that are key to any cell fate change using only CAGE-seq data as input.

Detection of new pioneer transcription factors as cell-type specific nucleosome binders

“…Several machine learning models have been proposed to predict transcription factor binding sites and identify sequence context features critical for TF binding 25,27–29 . In addition, gene regulatory network-based approaches have helped identify the key transcription factors in cell fate determination 30,31 . Despite the success of computational methods in genome-wide prediction of binding sites of canonical transcription factors, the locations of pioneer transcription factors’ binding sites, mechanisms of binding and regulation have not been systematically explored.…”

The neuronal homeobox transcription factor HMX3 is a crucial vulnerability factor in MECOM-negative KMT2A::MLLT3 acute myelomonocytic leukemia

“…Transcription factors that are key to cell fate change were determined using ANANSE (v0.4.0) 26 . TF binding profiles were predicted using CAGE-defined bidirectional regions, motif scores, and average ReMap ChIP-seq coverage, as described previously by Heuts et al 27 . Subsequently, gene regulatory networks (GRNs) were determined using summed unidirectional TCs per gene.…”

“…TFs that dictate leukemia cell fate through altered activity can be identified because they predominantly bind gene enhancers. By predicting genome-wide TF binding profiles in different cell types using enhancer activity and TF binding motifs, and by integrating these inferred binding profiles with genome-wide gene expression data, cell type-specific GRN and key TFs controlling cell fates can be identified 26,27 . Of all currently available bioinformatic algorithms, the Analysis Algorithm for Networks Specified by Enhancers (ANANSE) performs best in identifying cell fate determining TFs 26 .…”

“…For the latter, a proxy for open chromatin or histone H3K27-acetylation data are often used. We recently adapted the ANANSE approach by using solely 5' Cap Analysis of Gene Expression sequencing (CAGE-seq) data as input (ANANSE-CAGE) 27 . The advantage of CAGE-seq is that both expression of protein coding genes based on unidirectional expression as well as enhancer activity based on bidirectional expression of enhancer RNAs (eRNAs) can be determined in one go.…”

Detection of new pioneer transcription factors as cell-type-specific nucleosome binders

“…Several machine learning models have been proposed to predict TF binding sites and identify sequence context features critical for TF binding ( Sherwood et al, 2014 ; Zheng et al, 2021 ; Avsec et al, 2021 ; Kishan et al, 2021 ). In addition, gene regulatory network-based approaches have helped to identify the key TFs in cell fate determination ( Xu et al, 2021 ; Heuts et al, 2022 ). Despite the success of computational methods in genome-wide prediction of binding sites of canonical TFs, the locations of PTFs’ binding sites, mechanisms of binding, and regulation have not been systematically explored.…”