Artificial neural networks enable genome-scale simulations of intracellular signaling
Mammalian cells adapt their functional state in response to external signals in form of ligands that bind receptors on the cell-surface. Mechanistically, this involves signal-processing through a complex network of molecular interactions that govern transcription factor (TF) activity patterns. Computer simulations of the information flow through this network could help predict cellular responses in health and disease. Here we develop a recurrent neural network constrained by prior knowledge of the signaling network with ligand concentrations as input, TF activity as output and signaling molecules as hidden nodes. Simulations are assumed to reach steady state, and we regularize the parameters to enforce this. Using synthetic data, we train models that generalize to unseen data and predict the effects of gene knockouts. We also fit models to a small experimental data set from literature and confirm the predictions using cross validation. This demonstrates the feasibility of simulating intracellular signaling at the genome-scale.