Join us for a series of BCS Grdauate Student Research Talks on Friday, 3/11/22 at 4pm on Zoom https://mit.zoom.us/s/97547172548

Katya Tsimring, Year 4 Graduate Student in Mriganka Sur’s Lab

Title: Alignment of visual features in binocular cortical circuits through experience-dependent plasticity.

Abstract: Experience-dependent synaptic plasticity is essential for fine-tuning immature sensory circuits during critical periods in development. In visual circuits, this fine-tuning includes the alignment of information from the ipsilateral and contralateral eye onto neurons in the binocular primary visual cortex (bV1). At the onset of the critical period for binocular vision, bV1 neurons have distinct orientation preferences for each eye, which become matched by the end of the critical period. While we know that synapses on these neurons are highly plastic during the binocular critical period, it remains unclear how this plasticity leads to properly aligned bV1 neurons. Hebbian and heterosynaptic plasticity are mechanisms of experience-dependent plasticity that modify synaptic inputs based on the correlation of pre- and postsynaptic neuronal responses and on the activity of neighboring synapses, respectively. We propose that Hebbian and heterosynaptic mechanisms cooperatively regulate the binocular alignment of bV1 neurons during the critical period by strengthening synapses that are correlated with the postsynaptic neuron or with synaptic neighbors, and by weakening those that are uncorrelated. To examine the role these two mechanisms may play in rearranging binocular circuits over development, our ongoing research involves: 1) chronically tracking the eye-specific responses of bV1 neurons and their dendritic spines, and 2) mapping the synaptic composition of physiologically characterized dendritic spines post hoc. Our preliminary results suggest that ipsilateral inputs become aligned to soma’s orientation preference through the selective pruning of unmatched inputs. Together, our experiments aim to answer fundamental questions on the nature of experience-dependent plasticity in bV1 and provide critical insight into the synaptic basis of amblyopia and neurodevelopmental disorders arising from synaptic dysfunction.

Mikhail Khona, Year 4 Graduate Student in Ila Fiete’s Lab

Title - Grown not built: How developmental constraints and dynamics shape neural circuit activity

Abstract: What is the origin of structured activity in certain brain regions? The current computational paradigm is to construct an ecologically relevant task that the brain region of interest is implicated in solving and study the responses of appropriately trained networks optimized to perform these tasks. This approach misses a critical step that biology takes which is that of development - brains must be grown, through genetically encoded rules, from the embryonic stage until the organism matures. This process happens largely in the absence of a task and places several constraints which can have a strong influence on the structure of activity in mature circuits.

In this talk, I will cover our work on two cortical circuits: the grid cell circuit in the medial entorhinal cortex (MEC) and the primate visual cortical hierarchy. 

The grid cell system is involved in spatial navigation and grid cells occur in discrete modules. These modular grid responses emerge along a strip of the MEC with smooth gradients in several biophysical neural properties, within days of eye opening in juvenile animals. I will show how continuous attractor networks equipped with gradients in connectivity spontaneously self-organize into discrete modules through a mechanism we term 'peak selection'. This mechanism makes novel predictions about grid cells which we then confirm by matching the sequence of grid period ratios in previously collected electrophysiological data.

The primate visual cortex is modular, parcellated into a hierarchy of visual areas (V1, V2, V3, etc.) that abut each other in the brain. The areas are each retinotopically organized with several characteristing higher-level organizational features --- a notable example is that the polar angle of the retinotopic maps alternates (is mirrored) across area boundaries. Here, we take a developmental approach, using simple bottom-up rules like synaptic growth and pruning to grow an architecture for visual cortex that captures key large scale topographic properties observed experimentally.

Altogether, these studies demonstrate that many features of brain organization may arise not by direct optimization on particular tasks but as a consequence of low-level biophysical rules which unroll a cascade of developmental processes that determine structure and connectivity.

Heather Kosakowski, Year 5 Graduate Student in Nancy Kanwisher’s and Rebecca Saxe’s Lab

Title:  Interrogating cortical function of the awake infant brain using functional magnetic resonance imaging (fMRI).

Abstract:  Philosophers and psychologists have long debated the relative roles of built-in structure versus learning in the development of the human mind. Yet it is only recently that evidence from whole-brain measurements of awake infants has become available to inform these debates. First, I will discuss a new custom infant coil we designed that enabled us to collect a higher quantity, and quality, of awake infant fMRI data. I will then present evidence that infants have face-, scene-, and body-selective responses in the fusiform face area (FFA), parahipoocampal place area (PPA), and extrastriate body area (EBA) of infant ventral temporal cortex (VTC), challenging the “proto-architecture” hypothesis. I will wrap up with a brief discussion about the use of infant fMRI as a tool to interrogate infant cognition. 

Jungsoo Kim, Year 5 Graduate Student in Steve Flavell’s Lab

Title: Internal state modulates brain-wide representations of behavior in C. elegans
 

Abstract: The brain constantly fluctuates among a wide range of internal states that modulate how sensory cues are processed to give rise to behavior. Recent studies have shown these states are broadly reflected in neural activity across many brain regions. In addition, moment-by-moment behavioral variables are also represented in neural activity across many brain regions. This gives rise to a view that neural representations of internal states and acute behavioral variables co-exist in most brain regions, but how internal states impact the neural encoding of behavior remains largely unclear.

In this study, we address this question using the nematode C. elegans, whose transparent body and well-defined neural circuits allow us to analyze brain-wide representations of behavior across a range of internal states. First, we engineered a microscope that can capture whole-brain activity along with comprehensive behavioral information of a freely-moving animal. Using custom software, we can extract calcium signals with an order-of-magnitude improvement in signal-to-noise ratio compared to previous systems. These neural datasets, combined with comprehensive behavioral quantification, allow us to map the intricate relationship between brain-wide neural activity and behavior over varying internal states. To determine how each neuron encodes specific behavioral features, we developed a flexible nonlinear encoding model that can fit almost all of the neurons carrying overt behavioral information. Analysis of the resulting models reveals that neurons can encode the worm’s behavior over a wide range of timescales, encode multiple behaviors simultaneously in a nonlinear yet stereotyped fashion, and modulate their encoding of behavior depending on the animal’s internal state. Our results provide a global view of how circuits across the brain encode each motor parameter of an animal, and how changing internal states flexibly modulate these encodings.

Martin Schrimpf, PhD Year 5 Graduate Student in Jim DiCarlo’s Lab

Title: Advancing System Models of Brain Processing via Integrative Benchmarking

 Abstract: Research in the brain and cognitive sciences attempts to uncover the neural mechanisms underlying intelligent behavior in domains such as vision or language. Due to the complexities of brain processing, studies necessarily had to start with a narrow scope of experimental investigation and computational modeling. We argue that it is time for our field to take the next step: buildsystem models that capture neural mechanisms and supported behaviors within an entire domain of intelligence. To make progress on system models, we propose integrative benchmarking – integrating experimental results from many laboratories into suites of benchmarks that guide and constrain those models at multiple stages and scales. We show-case this approach by developing Brain-Score benchmark suites for neural and behavioral experiments in the primate visual ventral stream as well as the human language system. By systematically evaluating a wide variety of model candidates, we not only identify models beginning to match a range of brain data (~50% explained variance), but also discover key relationships: Models’ brain scores are predicted by their object categorization performance in vision (but only up to 70% ImageNet accuracy), and their next-word prediction performance in language. The better models predict internal neural activity, the better they match human behavioral outputs, with architecture substantially contributing to brain-like representations. Using the integrative benchmarks, we develop improved state-of-the-art system models that more closely match shallow recurrent neuroanatomy and predict primate temporal processing. Taken together, these integrative benchmarks and system models are first steps to modeling the complexities of brain processing in entire domains of intelligence.