Modeling gene transcription

National Institute Of Diabetes And Digestive And Kidney Diseases

Investigators

Linked publications, trials & patents

Abstract

Previously, my group developed a biochemical model for steroid-mediated gene expression in the presence of various factors. Experiments have found that the dose-response curve for gene expression closely follows a Michaelis-Menten function and that factors can alter both the maximum value and location of half maximum of the function. We showed theoretically that this highly stringent constraint can only occur in a sequence of reactions if factors downstream of receptor-steroid binding interact weakly. The theory can then make precise predictions on the mechanisms and site of action of these cofactors. We used the theory to design a novel competition assay to predict the mechanisms and relative positions of the two cofactors and have applied it to several different factors including the cancer related gene MYC. I now collaborate with experimentalists to attempt to reconcile the theory with single cell imaging data. To do so, I have developed a stochastic single gene model that is consistent with the previous model. Instead of just predicting the amount of gene product, the model predicts the time kinetics of transcription and the amount of mRNA produced, which can be compared to live cell real time single RNA molecule imaging and mRNA smFISH data. We have shown that gene transcription has kinetics on multiple time scales. Genes generally transcribe by actively bursting for random intervals on the order of tens of minutes but can go into deeply repressed states for days or weeks between bursts. Gene splicing also occurs stochastically and recursively. I have also applied the model to scRNA data from the entire genome. To facilitate fitting of model to data I have written an open source software suite called StochasticGene.jl.