An Impulse model for temporal response profiles of gene expression

Cellular responses to environmental changes are often characterized by typical temporal response profiles of their gene expression. Modeling these response profiles can help us understand how the transcriptional program of the cell changes in time to cope with environmental perturbations. We describe here a parametric model designed to capture the characteristic temporal profiles. This impulse model, explicitly represents natural temporal properties of the responses such as the onset and the offset time. It is simple enough to be learned from sparse data, and can robustly fit a temporal pattern even to the expression data of a single gene. We evaluate our model on the task of filling missing values in gene expression data, showing that it significantly outperforms all current methods, both at the level of individual genes, and when allowed to use whole genome information. We further show how the functionally relevant representation of the expression time-profiles, has direct biological interpretation, and provides insight about the cell’s response to environmental perturbations, and demostrate it with application to cell memory after heat shock.