A general recurrent neural network approach to model genetic regulatory networks

There is an urgent need for tools to unravel the complex interactions and functionalities of genes. As such, there has been much interest in reverse-engineering genetic regulatory networks from time series gene expression data. We use an artificial neural network to model the dynamics of complicated gene networks and to learn their parameters. The positive and negative regulations of genes are defined by a weight matrix, and different genes are allowed to have different decaying time constants. We demonstrate the effectiveness of the method by recreating the SOS DNA Repair network of Escherichia coli bacterium, previously discovered through experimental data.