Partnering to Accelerate Neural Network Models

Sep 21, 2021 | Tech Life


In the lab of Günther K.H. Zupanc, the Brown Ghost Knifefish (Apteronotus leptorhynchus) has an important role in advancing Zupanc’s research into human neurological systems. The small fish, common to the Amazon basin, has an electric surprise within its body. Part of a species that have electric organs that can produce electric fields, the knifefish’s weak electric discharge has properties like neural oscillators in human brains. These oscillators in humans are what control our rhythmic behaviors, such as breathing and heartbeat.

Modeling Neural Networks on the Discovery Cluster

By building a computational model of an oscillator called a pacemaker nucleus, Zupanc was able to find strong theoretical evidence of how certain ions play a role in modulating the frequency of the neural oscillations that the pacemaker nucleus produces.

However, running these types of very complex models takes a lot of computational power that can take weeks or even months to run on the most advanced desktop system. In order to reduce compute time, Zupanc’s team turned to the Northeastern’s Research Computing Team (RC) for assistance to access the Discovery cluster, the high performance computing resource for the Northeastern University research community.

RC’s Computational Scientist  

Mariana Levi, Ph.D. joined Northeastern’s Research Computing team in 2019. Before joining the team, she was a Ph.D. student in the Whitford Research Group, focused on theoretical and computational molecular biophysics. Currently, Mariana is the Research Computing team’s Computational Scientist, assisting researchers with a variety of assistance from help with installing scientific software programs, improving the efficiency and speed of their code, and leading training sessions on creating resilient workflows.

She was first approached by Daniel Hartman, an undergraduate researcher in Gunther’s lab, with accessing and using the Discovery Cluster.  “They probably couldn’t have done it on their personal laptop, “ Levi said. “It was just impossible. It definitely was something made possible with Discovery.”

Collaboration Accelerating Science

Working closely with Zupanc’s team, Levi was able to provide specialized assistance to help build and refine computational models. “Mariana has a strong research background in biophysics,” Zupanc states, “She not only could help with adjusting the computer codes for running the simulations on the Discovery, but also was able to understand the research project and provide her input to the study.” Levi’s input was so critical to the success of the study, she was given co-authorship on the resulting paper, Modeling of sustained spontaneous network oscillations of a sexually dimorphic brainstem nucleus: the role of potassium equilibrium potential, published in 2021 in the Journal of Computational Neuroscience.

“One of the most positive aspects of this work was the great team spirit among all members of the team, including Mariana as our Research Computing collaborator. A complex and challenging project such as the one we did is possible only if all the team members work together very closely and very well. Mariana is an incarnation of such a team spirit!” –Gunther Zupanc

How can Northeastern’s Research Computing team help accelerate your research?

Research Computing is here to serve the computational and storage needs of the Northeastern University research community. No request is too small, and the team is also delighted to work with researchers on longer-term, complex issues to help fast-track your research to publication. Reach out to Research Computing by emailing or schedule a consultation on the RC Bookings page to get started today!



Hartman, Daniel, Lehotzky, David, Ilies, Iulian, Levi, Mariana, & Zupanc, Gunther K. H. (2021). Modeling of sustained spontaneous network oscillations of a sexually dimorphic brainstem nucleus: the role of potassium equilibrium potential. Journal of Computational Neuroscience.