Fig 1.
(A) A toy model of gene regulation of 3 genes involved in a transcriptional regulatory network, showing genes transcribed into mRNAs and translated into proteins that regulate another one of the genes. (B) Compressed representation of the interactions on the left as a GRN in which only genes are shown with their regulatory interactions.
Fig 2.
From time-series transcriptomics to GRN inference and model validation.
The process usually starts with processing the time-series datasets and identifying the time-points that are relevant to describe the biological process (1). As a second step, data formatting, normalization and gene filtering and/or binarization might be necessary (2), depending on the inference method to be used (3). The inferred network consists of a weighted directed GRN (4). Model validation (5) is then necessary to quantify the method performance in inferring the GRN when compared to gold-standard datasets, evidence from literature or prior biological knowledge. The inferred network can further be used for the application of dynamical models in cellular or multilevel modeling (6).
Table 1.
GRN inference tools from time-series transcriptomics categorized by their inferring algorithm.
The characteristics of the inferred network are indicated as follows: ⊘ undirected; ⊳ directed and unsigned; ▶ directed and signed.
Fig 3.
GRN inference from single-cell (pseudo)time-series.
Starting from single-cell (pseudo) time-series (1), the inference… workflow follows a similar path as in bulk time-series transcriptomics, except from additional steps of dimensionality reduction and trajectory inference (2), and pseudo-time ordering of the cells (3) when time-resolved experimental measurements are not available. Steps (4–6) follow the same logic as in Fig 2.
Table 2.
GRN inference tools from single-cell transcriptomics categorized by their inferring algorithm.
The characteristics of the inferred network are indicated as follows: ⊘ undirected; ⊳ directed and unsigned; ▶ directed and signed.
Fig 4.
(A) Dynamical models of gene regulatory networks, represented as a spectrum of models ranging from continuous to discrete. (B) Steady states of a system composed of 3 genes: (left) ODE model, as a system of 3 interacting species; (right) Boolean model, as logic-based interacting entities. The attractors are identified as steady states in the long-term behavior of the system.