Modeling Signaling Pathways Using Recurrent Neural Networks (DSI-SRP)
This DSI-SRP fellowship funded Alexander Lin to work in the laboratory of Dr. Gregor Neuert, Ph.D. in the Department of Molecular Physiology and Biophysics during the summer of 2023. Alexander is a rising sophomore with majors in Computer Science and Biochemistry and a minor in Data Science.
His project focused on the Hog1 signaling network, an evolutionarily conserved pathway found in yeast that is crucial in the organism’s ability to adapt to environmental perturbations. Alexander and his mentor demonstrated that a simplified dynamic ordinary differential equation (ODE) model of this pathway can accurately predict the temporal signaling responses that result from time-varying changes in environmental stimuli. Here, they devised and implemented a customized, interpretable recurrent neural network (RNN) model to achieve a similar goal. Specifically, they constrained this RNN model based on identified Hog1 pathway interactions from the literature – including activation, deactivation, and negative feedback regulations. They implemented these regulations through a tailored RNN architecture and trained these RNN models using experimentally measured Hog1 activation data. These models were able to accurately predict pathway activation under new stimulation conditions. Developing such predictive models of signaling networks will allow researchers to computationally forecast the result of protein mutations and gain insight into the relationships between protein-protein interactions within the signaling network.
In addition to receiving support through a DSI-SRP fellowship, this project was supported and facilitated by the DSI Data Science Team through their regular summer workshops and demo sessions.