A neural network-based model framework for cell-fate decisions and development

Mátyás Paczkó,Dániel Vörös,Péter Szabó,Gáspár Jékely,Eörs Szathmáry,András Szilágyi
DOI: https://doi.org/10.1038/s42003-024-05985-1
IF: 6.548
2024-03-15
Communications Biology
Abstract:Gene regulatory networks (GRNs) fulfill the essential function of maintaining the stability of cellular differentiation states by sustaining lineage-specific gene expression, while driving the progression of development. However, accounting for the relative stability of intermediate differentiation stages and their divergent trajectories remains a major challenge for models of developmental biology. Here, we develop an empirical data-based associative GRN model (AGRN) in which regulatory networks store multilineage stage-specific gene expression profiles as associative memory patterns. These networks are capable of responding to multiple instructive signals and, depending on signal timing and identity, can dynamically drive the differentiation of multipotent cells toward different cell state attractors. The AGRN dynamics can thus generate diverse lineage-committed cell populations in a robust yet flexible manner, providing an attractor-based explanation for signal-driven cell fate decisions during differentiation and offering a readily generalizable modelling tool that can be applied to a wide variety of cell specification systems.
biology
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is that the existing models of cell fate determination and development processes are insufficient in explaining the relative stability of intermediate differentiation stages and their different trajectories. Specifically, although a large number of theoretical and experimental studies have focused on lineage selection, the regulatory roles of the functional building blocks of gene regulatory networks (GRNs) in cell fate determination have not been systematically and fully mapped. Therefore, how to define cell states and the transitions between them, and how environmental cues, cell - intrinsic mechanisms and their interactions control these processes remain unsolved mysteries. To meet this challenge, the authors developed an associative GRN model (AGRN) based on empirical data, which can store multi - lineage - stage - specific gene expression profiles as associative memory patterns. Through this model, pluripotent cells can be dynamically driven to differentiate towards different cell - state attractors, thereby generating diverse lineage - committed cell populations in a robust and flexible manner. The dynamics of AGRN provides an attractor - based explanation for the signal - driven cell fate determination process, and this model can be applied to a wide range of cell - specialization systems. ### Main contributions 1. **Model framework**: Proposed a neural - network - based model framework for cell fate determination and development processes. 2. **Associative memory**: Store multi - lineage - stage - specific gene expression profiles through associative memory patterns. 3. **Dynamic response**: The model can respond to multiple guiding signals and dynamically drive pluripotent cells to differentiate towards different cell - state attractors according to the time and identity of the signals. 4. **Attractor explanation**: Provide an attractor - based explanation for the signal - driven cell fate determination process. 5. **Wide applicability**: The model can be applied to multiple cell - specialization systems and has high universality and flexibility. ### Specific implementation - **Gene regulatory network**: Gene regulatory networks (GRNs) maintain the stability of cell differentiation states through activation and inhibition interactions while promoting the development process. - **Associative memory network**: Using the principles of associative memory networks, gene expression patterns are regarded as attractors, and each attractor corresponds to a specific cell differentiation stage. - **Dynamics model**: By constructing a regulatory matrix \( M \), the model can dynamically regulate gene expression levels, thereby describing the changes in gene expression during cell differentiation. ### Mathematical framework - **Gene expression vector**: The state of the system at time \( t \) can be represented by the gene expression vector \( \mathbf{p}(t)=(p_1, p_2, \ldots, p_N)^T \), where \( p_i \) represents the expression level of the \( i \) - th gene. - **Regulatory matrix**: The element \( m_{ij} \) in the regulatory matrix \( M \) defines the regulatory effect, and positive and negative values represent activation and inhibition effects respectively. - **Dynamics equation**: The change in gene expression can be described by the following differential equation: \[ \frac{d\mathbf{p}(t)}{dt}=f(M \mathbf{p}(t)) \] where \( f \) is a nonlinear activation function, and a scaled Sigmoid function is usually used. ### Results verification - **Hematopoietic system**: The model successfully reproduced the experimentally verified dynamic gene expression in the human hematopoietic system. - **Cell cycle**: The model can accurately describe the cyclic dynamics of the cell cycle and can terminate the cyclic dynamics through external signals. - **Caenorhabditis elegans embryonic development**: The model successfully described the changes in gene expression during the embryonic development of Caenorhabditis elegans, and the performance did not decrease significantly even when dealing with large - scale data sets. ### Robustness analysis - **Perturbation analysis**: The model shows strong robustness to multiplicative and nullifying perturbations of regulatory intensity, and most simulations can still correctly pass through the predetermined paths even under large perturbations. Through these methods and results, this paper provides a new perspective for understanding the gene regulatory mechanisms in cell fate determination and development processes.