Gene expression can be used to determine whether a cell is healthy or diseased. This gene expression depends on the cell’s state, which is influenced by its previous state. We can categorize the cell’s state into the source state and the target state. The changes between the source and target states occur because of the activation and repression of some of the genes present in the cell. The identification of the genes that drive the transformation from the source state to the target state is important, as this transformation enables early detection of diseases like cancer and tumors. However, identifying these genes remains an unresolved challenge.
We can represent the complex relationships between genes and transcription factors that regulate and determine the behavior of the cell using a gene regulatory network (GRN). In GRNs, the nodes represent genes, and the edges represent the regulatory relationships between genes. Accurate identification of target genes requires knowledge of the GRN in both the source and target cell states. GRNs are extremely important as they help us understand the biological processes within the cell. One key application of GRNs is the comparison of healthy versus diseased tissues. Such studies help us identify genes that may be responsible for diseases.
Some approaches to inferring the relationship between the source state and target state include differential expression analysis, machine learning and deep learning-based perturbation algorithmic approaches, or manual perturbation of specific transcription factors (TFs). In the paper GAGER (GRN Assisted Gene Expression Restoration), the authors designed an algorithm to identify specific genes whose manipulation transforms the source state into the target state of the cell by comparing the GRNs of the source and target cell states. The algorithm mainly applies a series of perturbations to convert the source cell state into the target cell state. It aims to infer the differences between the regulatory edges of the GRNs of the source and target cell states. From this, it identify transcription factors that potentially change the cell state from source to target. One limitation of their work is the assumption of a simple linear relationship between genes, while the relationship between genes may actually be more complex. They try to perturb TFs greedily to reduce the expression differences between the source and target cell states. We aim to explore gene expression restoration further by assuming that the relationships between genes are complex.