Predictive models of cell state change in health and disease

Looking at an ensemble of states of an Ising model, we can train restricted Boltzmann machines to learn the corresponding probability distribution. Once this network is trained, there is a way to infer the actual couplings between Ising sites. In my project, we hope to generalise this technique to apply it to gene expression levels, and use similarly trained networks to infer the regulation couplings among the genes. From these networks, we then hope to be able to formulate some rules on the behaviour of the network and its relation to the eventual state, or phenotype, of the cell.