Biology subjects

Natrajan, M.

Publications and source records attributed to Natrajan, M..

4 recordsLinked to original sources

bioRxiv2024

Stability through plasticity: Finding robust memories through representational drift

Memories are believed to be stored in synapses and retrieved through the reactivation of neural ensembles. Learning alters synaptic weights, which can interfere with previously stored memories that share the same synapses, creating a tradeoff between plasticity and stability. Interestingly, neural representations exhibit significant dynamics, even in stable environments, without apparent learning or forgetting--a phenomenon known as representational drift. Theoretical studies have suggested that multiple neural representations can correspond to a memory, with post-learning exploration of these representation solutions driving drift. However, it remains unclear whether representations explored through drift differ from those learned or offer unique advantages. Here we show that representational drift uncovers noise-robust representations that are otherwise difficult to learn. We first define the non-linear solution space manifold of synaptic weights for a fixed input-output mapping, which allows us to disentangle drift from learning and forgetting and simulate representational drift as diffusion within this manifold. Solutions explored by drift have many inactive and saturated neurons, making them robust to weight perturbations due to noise or continual learning. Such solutions are prevalent and entropically favored by drift, but their lack of gradients makes them difficult to learn and non-conducive to further learning. To overcome this, we introduce an allocation procedure that selectively shifts representations for new information into a learning-conducive regime. By combining allocation with drift, we resolve the tradeoff between learnability and robustness.

neuroscience↗

bioRxiv2023

Learning produces a hippocampal cognitive map in the form of an orthogonalized state machine

Cognitive maps confer animals with flexible intelligence by representing spatial, temporal, and abstract relationships that can be used to shape thought, planning, and behavior. Cognitive maps have been observed in the hippocampus, but their algorithmic form and the processes by which they are learned remain obscure. Here, we employed large-scale, longitudinal two-photon calcium imaging to record activity from thousands of neurons in the CA1 region of the hippocampus while mice learned to efficiently collect rewards from two subtly different versions of linear tracks in virtual reality. The results provide a detailed view of the formation of a cognitive map in the hippocampus. Throughout learning, both the animal behavior and hippocampal neural activity progressed through multiple intermediate stages, gradually revealing improved task representation that mirrored improved behavioral efficiency. The learning process led to progressive decorrelations in initially similar hippocampal neural activity within and across tracks, ultimately resulting in orthogonalized representations resembling a state machine capturing the inherent structure of the task. We show that a Hidden Markov Model (HMM) and a biologically plausible recurrent neural network trained using Hebbian learning can both capture core aspects of the learning dynamics and the orthogonalized representational structure in neural activity. In contrast, we show that gradient-based learning of sequence models such as Long Short-Term Memory networks (LSTMs) and Transformers do not naturally produce such orthogonalized representations. We further demonstrate that mice exhibited adaptive behavior in novel task settings, with neural activity reflecting flexible deployment of the state machine. These findings shed light on the mathematical form of cognitive maps, the learning rules that sculpt them, and the algorithms that promote adaptive behavior in animals. The work thus charts a course toward a deeper understanding of biological intelligence and offers insights toward developing more robust learning algorithms in artificial intelligence.

neuroscience↗

bioRxiv2022

Leveraging place field repetition to understand positional versus nonpositional inputs to hippocampal field CA1

The hippocampus is believed to encode episodic memory by binding information about the content of experience within a spatiotemporal framework encoding the location and temporal context of that experience. Previous work implies a distinction between positional inputs to the hippocampus that provide information about an animals location and nonpositional inputs which provide information about the content of experience, both sensory and navigational. Here we leverage the phenomenon of "place field repetition" to better understand the functional dissociation between positional and nonpositional inputs to CA1. Rats navigated freely on a novel maze consisting of linear segments arranged in a rectilinear, city-block configuration, which combined elements of open-field foraging and linear-track tasks. Unlike typical results in open-field foraging, place fields were directionally tuned on the maze, even though the animals behavior was not constrained to 1-D trajectories. Repeating fields from the same cell tended to have the same directional preference when the fields were aligned along a linear corridor of the maze, but they showed uncorrelated directional preferences when they were unaligned across different corridors. Lastly, individual fields displayed complex time dynamics which resulted in the population activity changing gradually over the course of minutes. These temporal dynamics were evident across repeating fields of the same cell. These results demonstrate that the positional inputs that drive a cell to fire in similar locations across the maze can be behaviorally and temporally dissociated from the nonpositional inputs that alter the firing rates of the cell within its place fields, thereby increasing the flexibility of the system to encode episodic variables within a spatiotemporal framework provided by place cells.

neuroscience↗

bioRxiv2020

Alterations in the human plasma lipidome in response to Tularemia vaccination

Tularemia is a rare but highly contagious and potentially fatal disease caused by bacteria Francisella tularensis where as few as ten inhaled organisms can lead to an infection, making it one of the most infectious microorganisms known and a potential bioweapon. To better understand the response to a live, attenuated tularemia vaccine and the biological pathways altered post-vaccination, healthy adults were vaccinated by scarification and plasma was collected pre- and post-vaccination for longitudinal lipidomics studies. Using tandem mass spectrometry, we identified and quantified individual lipid molecular species within representative lipid classes in plasma to characterize alterations in the plasma lipidome during the vaccine response. Separately, we targeted oxylipins, a subset of lipid mediators involved in inflammatory pathways. We identified 14 differentially abundant lipid species from eight lipid classes. These included 5-Hydroxyeicosatetraenoic acid (5-HETE), an eicosanoid produced following arachidonic acid liberation and epoxygenation, which is indicative of lipoxygenase activity and, subsequently, inflammation. Results suggest that 5-HETE was metabolized to a dihydroxyeicosatrienoic acid (DHET) by Day 7 post-vaccination, shedding light on the kinetics of the 5-HETE-mediated inflammatory response. In addition to 5-HETE and DHET, we observed pronounced changes in 34:1 phosphatidylinositol, anandamide, oleamide, ceramides, 16:1 cholesteryl ester, and several glycerophospholipids, several of these changes in abundance were correlated with serum cytokines and T cell activation. These data provide new insights into alterations in plasma lipidome post tularemia vaccination, potentially identifying key mediators and pathways involved in vaccine response and efficacy.

biochemistry↗