scholarly journals Data-driven modeling predicts gene regulatory network dynamics during the differentiation of multipotential hematopoietic progenitors

2022 ◽  
Vol 18 (1) ◽  
pp. e1009779
Author(s):  
Joanna E. Handzlik ◽  
Manu

Cellular differentiation during hematopoiesis is guided by gene regulatory networks (GRNs) comprising transcription factors (TFs) and the effectors of cytokine signaling. Based largely on analyses conducted at steady state, these GRNs are thought to be organized as a hierarchy of bistable switches, with antagonism between Gata1 and PU.1 driving red- and white-blood cell differentiation. Here, we utilize transient gene expression patterns to infer the genetic architecture—the type and strength of regulatory interconnections—and dynamics of a twelve-gene GRN including key TFs and cytokine receptors. We trained gene circuits, dynamical models that learn genetic architecture, on high temporal-resolution gene-expression data from the differentiation of an inducible cell line into erythrocytes and neutrophils. The model is able to predict the consequences of gene knockout, knockdown, and overexpression experiments and the inferred interconnections are largely consistent with prior empirical evidence. The inferred genetic architecture is densely interconnected rather than hierarchical, featuring extensive cross-antagonism between genes from alternative lineages and positive feedback from cytokine receptors. The analysis of the dynamics of gene regulation in the model reveals that PU.1 is one of the last genes to be upregulated in neutrophil conditions and that the upregulation of PU.1 and other neutrophil genes is driven by Cebpa and Gfi1 instead. This model inference is confirmed in an independent single-cell RNA-Seq dataset from mouse bone marrow in which Cebpa and Gfi1 expression precedes the neutrophil-specific upregulation of PU.1 during differentiation. These results demonstrate that full PU.1 upregulation during neutrophil development involves regulatory influences extrinsic to the Gata1-PU.1 bistable switch. Furthermore, although there is extensive cross-antagonism between erythroid and neutrophil genes, it does not have a hierarchical structure. More generally, we show that the combination of high-resolution time series data and data-driven dynamical modeling can uncover the dynamics and causality of developmental events that might otherwise be obscured.

2021 ◽  
Author(s):  
Joanna Elzbieta Handzlik ◽  
Manu Manu

Cellular differentiation during hematopoiesis is guided by gene regulatory networks (GRNs) thought to be organized as a hierarchy of bistable switches, with antagonism between Gata1 and PU.1 driving red- and white-blood cell differentiation. We utilized high temporal-resolution gene-expression data from in vitro erythrocyte-neutrophil differentiation and a predictive data-driven dynamical modeling framework to learn the architecture and dynamics of gene regulation in a comprehensive twelve-gene GRN. The inferred genetic architecture is densely interconnected rather than hierarchical. The analysis of model dynamics revealed that neutrophil differentiation is driven by C/EBPα and Gfi1 rather than PU.1. This prediction is supported by the sequence of gene upregulation in an independent mouse bone marrow scRNA-Seq dataset. These results imply that neutrophil differentiation is not driven by the Gata1-PU.1 switch but by neutrophil-specific genes instead. More generally, this work shows that data-driven dynamical modeling can uncover the causality of developmental events that might otherwise be obscured.


2020 ◽  
Vol 36 (19) ◽  
pp. 4885-4893 ◽  
Author(s):  
Baoshan Ma ◽  
Mingkun Fang ◽  
Xiangtian Jiao

Abstract Motivation Gene regulatory networks (GRNs) capture the regulatory interactions between genes, resulting from the fundamental biological process of transcription and translation. In some cases, the topology of GRNs is not known, and has to be inferred from gene expression data. Most of the existing GRNs reconstruction algorithms are either applied to time-series data or steady-state data. Although time-series data include more information about the system dynamics, steady-state data imply stability of the underlying regulatory networks. Results In this article, we propose a method for inferring GRNs from time-series and steady-state data jointly. We make use of a non-linear ordinary differential equations framework to model dynamic gene regulation and an importance measurement strategy to infer all putative regulatory links efficiently. The proposed method is evaluated extensively on the artificial DREAM4 dataset and two real gene expression datasets of yeast and Escherichia coli. Based on public benchmark datasets, the proposed method outperforms other popular inference algorithms in terms of overall score. By comparing the performance on the datasets with different scales, the results show that our method still keeps good robustness and accuracy at a low computational complexity. Availability and implementation The proposed method is written in the Python language, and is available at: https://github.com/lab319/GRNs_nonlinear_ODEs Supplementary information Supplementary data are available at Bioinformatics online.


2021 ◽  
Vol 30 (04) ◽  
pp. 2150022
Author(s):  
Sergio Peignier ◽  
Pauline Schmitt ◽  
Federica Calevro

Inferring Gene Regulatory Networks from high-throughput gene expression data is a challenging problem, addressed by the systems biology community. Most approaches that aim at unraveling the gene regulation mechanisms in a data-driven way, analyze gene expression datasets to score potential regulatory links between transcription factors and target genes. So far, three major families of approaches have been proposed to score regulatory links. These methods rely respectively on correlation measures, mutual information metrics, and regression algorithms. In this paper we present a new family of data-driven inference methods. This new family, inspired by the regression-based paradigm, relies on the use of classification algorithms. This paper assesses and advocates for the use of this paradigm as a new promising approach to infer gene regulatory networks. Indeed, the development and assessment of five new inference methods based on well-known classification algorithms shows that the classification-based inference family exhibits good results when compared to well-established paradigms.


2017 ◽  
Author(s):  
Xiaona Fang ◽  
Qiong Liu ◽  
Christopher Bohrer ◽  
Zach Hensel ◽  
Wei Han ◽  
...  

AbstractBistable switches are common gene regulatory motifs directing two mutually exclusive gene expression states, and consequently distinct cell fates. Theoretical studies suggest that the simple circuitry of bistable switches is sufficient to encode more than two cell fates due to the non-equilibrium, heterogeneous cellular environment, allowing a high degree of adaptation and differentiation. However, new cell fates arising from a classic bistable switch without rewiring the circuitry have not been experimentally observed. By developing a new, dual single-molecule gene-expression reporting system in liveE. colicells, we investigated the expression dynamics of two mutually repressing transcription factors, CI and Cro, in the classic genetic switch of bacteriophage λ. We found that in addition to the two expected high-Cro and high-CI production states, there existed two new ones, in which neither CI nor Cro was produced, or both CI and Cro were produced. We constructed the corresponding potential landscape and mapped the transition kinetics between the four production states, providing insight into possible state-switching rates and paths. These findings uncover new cell fate potentials beyond the classical picture of λ switch, and open a new window to explore the genetic and environmental origins of the cell fate decision-making process in gene regulatory networks.


Sign in / Sign up

Export Citation Format

Share Document