ANN-based simulation of transcriptional networks in Yeast

Artificial Neural Networks (ANNs) have recently been used to quantitatively model stress-induced transcriptional changes in gene regulatory networks, based on gene expression and DNA-binding information. Here, we extend this approach to study the MSN2/4 regulatory networks in Yeast, which are known to be involved in the stress response. We also refine the ANN models by incorporating the dynamics of transcriptional regulation and test our method on three networks involving YAP1. For the latter we make an extensive search in order to identify potential latencies between transcriptional activation and corresponding changes in the expression of targeted genes. Finally, we test our model´s ability to replicate gene-deletion findings in the YAP1 networks. We find that our models can accurately capture the regulatory effect of different transcription factors, under both normal and gene knockout conditions and that incorporating latencies in the ANN models results in significantly higher performance. Overall, we show that ANNs can be used to provide quantitative predictions about the expression profile of targeted genes during the stress response in Yeast.