Skip to main content
Figure 3 | BMC Systems Biology

Figure 3

From: The extended TILAR approach: a novel tool for dynamic modeling of the transcription factor network regulating the adaption to in vitro cultivation of murine hepatocytes

Figure 3

Concept of modeling of TILAR (A-D) compared to ExTILAR (E-G). A-D) The modeling concept of the TILAR algorithm. The genes are labeled with their expression values. A) In TILAR, a gene can only be regulated by another gene via a TF k if the regulating gene does not possess a TFBS for the TF k itself (TF-to-gene realtions). B) This decreases the number of possible network topologies and therefore serves as a additional source of prior-knowledge (gene-to-TF relations). C) LARS is used to infer a sparse network which explains the measured expression values of the genes in the best possible way. A constrained ordinary least square (OLS) approach is used to estimate the parameters using the final structure obtained from LARS. D) This way, new hypotheses about gene to gene relations can be obtained. E-G) The extended concept of modeling used by ExTILAR. Since the algorithm estimates the change of expression of each gene over time, the nodes are labeled with ŷ i = Δ x i Δt where Δ x i =x i [t m ]−x i [tm−1] and Δt=1 is outlined in the labels of the corresponding genes. E) The number of possible network structures is lowered by the TFBS information. Additionally, auto-regulation and modeling input perturbations are introduced and increase the number of regression coefficients to estimate. F) One possible model is selected from the full set of models returned by LARS. A OLS approach is used to find the parameters, given the network structure of the selected model. G) The gene expression dynamics of the final network can be simulated using standard ODE-solvers.

Back to article page