Fereshteh Lagzi, Ph.D.

Fereshteh Lagzi is a postdoctoral fellow in the Department of Physiology and Biophysics at the University of Washington (UW). She has a background in computational neuroscience, nonlinear dynamics and control theory. During her graduate research., Lagzi studied the dynamics of competition and cooperation in spiking networks of interacting subnetworks, and their bifurcation. After completing her Ph.D., she studied artificial neural networks and analyzed different recurrent networks from a dynamical system point of view. In her postdoctoral research, she studied assembly formation considering cell-specific plasticity mechanisms for different inhibitory subtypes and their interaction with excitatory Hebbian learning mechanisms. Her studies link rate and weight dynamics in spiking networks in an interesting low-dimensional framework. Lagzi received the competitive Swartz Foundation Postdoctoral Fellowship in computational neuroscience and joined the Fairhall Lab in 2021. Currently, Lagzi studies circuit dynamics underlying sequence generation and maintenance. More precisely, she is interested in understanding the role of sequences in learning. Also, the dynamics and emergence of cell assemblies mediated by inhibitory subtypes are an active area of her research. Lagzi also considers alternative mechanisms to synaptic plasticity that can modify circuit dynamics and guide learning.

Principal Investigator: Adrienne Fairhall

Fellow: Priyanka Rao

Undergraduate Fellow Project:

Project 1: Many events unfold over time and/or space, giving both experiences and outputs such as language, thoughts and motor activity a sequential structure. Sequential patterns in neural activity have been observed in multiple brain areas, but most notably in the hippocampus, an area involved in learning and memory, where they recur spontaneously following experience. The acquisition of sequences may therefore be a fundamental mechanism the brain uses to learn and store relationships. In this project, we will use modeling and simulations to explore different mechanisms for building neural sequences. Many current models assume that sequences are learned by synaptic plasticity, i.e., strengthening connections between neurons such that driving the first neuron to fire can activate the entire sequence. However, recent work has brought attention to the potential general importance of nonsynaptic plasticity mechanisms. We will explore how changes in intrinsic neuronal properties, such as excitability, as well as the non-random structure of the network can contribute to sequence generation, and how the resulting models differ from more traditional synaptic approaches. The results of this study will highlight the necessary components for the generation and maintenance of sequences in neuronal networks.

Project 2: Typically in neural networks, neurons are modeled as connected with chemical synapses, whereby signals are transmitted in an all or none thresholded fashion at precise times. When we train artificial neural networks, we generally assume that learning occurs through changes in synaptic weights. However, many biological systems have neural connectivities that are mediated also or instead by gap junctions, which couple neurons electrically, leading to a continuous sharing of electrical potential. The motif of gap junctional coupling is widespread throughout neural systems; intriguingly, particularly during early development in cortex, in regions that undergo neurogenesis, such as the birdsong circuit nucleus HVC and in the highly regenerative organism Hydra. What role might this coupling play? We hypothesize that gap junctional coupling provides a mechanism for neurons to form initial connectivity, bringing new neurons into a coupled relationship that may subsequently be refined by synaptic plasticity. In this project we will explore this idea using modeling of Hydra and of HVC, where well-defined activity patterns (global repetitive firing of subnetworks and precise temporal sequences) have been well studied.