Genetic network inference via gene set stochastic sampling and sensitivity analysis

In this paper, two approaches utilizing neural networks, intended to infer genetic regulatory networks from temporal gene expression measurements, are examined. These approaches aimed to find a minimal set of genes that were able to accurately predict the expression levels of a given gene, thus modeling the interactions in the underlying genetic regulatory networks. Two neural network architectures were employed in each approach to determine the robustness of the modeling procedure with respect to the network architecture. Two testing procedures were also devised to evaluate the trained neural networks' performance and generalizability. The resulting neural networks predicted, with high accuracy, the target gene expression level at future times given the predicted minimal gene-set expression levels at previous time points