BCS Graduate Student Research Presentations for Interview Day
Description
BCS Graduate Student Research Presentations for Interview Day
Date: Friday, February 27th at 9AM
On Zoom: https://mit.zoom.us/j/99479394726
Presenter: Thomas Clark
Advisors: Ted Gibson & Roger Levy
Modeling Noisy-Channel Comprehension: Incremental Processing and Reanalysis as Probabilistic Inference
Abstract: How are comprehenders able to extract meaning from utterances in the presence of production errors? The Noisy-Channel theory provides an account grounded in Bayesian inference: comprehenders may interpret utterances non-literally when there is an alternative with higher prior probability and a plausible error likelihood. Yet the question of how such alternatives are generated and evaluated is open to debate. One obstacle has been the lack implemented computational models capable of predicting human processing of arbitrary utterances and handling the considerable uncertainty present in building a generative model of “noisy” language. Here, we model noisy-channel processing as approximate probabilistic inference over intended sentences and production errors. We combine Sequential Monte Carlo methods with “rejuvenation” moves to yield an algorithm that is incremental but allows reanalysis of previous material. We demonstrate evidence that the model reproduces established patterns of interpretations for sentences varying in plausibility and structure, and predicts patterns of re-reading behavior for anomalous sentences in a novel reading experiment. Finally, we extend our model to cases of mutual intelligibility between unrelated languages. Our results offer a step towards a flexible, algorithmic account of inference during real-world language comprehension.
______________________________________________________________________________
Presenter: Mingus Zoller
Advisor: Li-Huei Tsai
Neuronal DNA Damage and Hyperactivity Regulate MHC-I-Dependent T Cell Recruitment in the AD Brain
Abstract: Alterations in the adaptive immune system are now recognized as a key feature of Alzheimer's disease (AD). T cells have been observed in the brain parenchyma of humans with AD and in other neurodegenerative disorders. In AD mouse models of tau pathology and amyloidosis, T cells in the brain parenchyma have been shown to contribute to AD pathogenesis. However, the precise molecular mechanisms governing T cell recruitment to the brain and their subsequent role in neurodegeneration remain to be elucidated.
Using Focused Ion Beam - Scanning Electron Microscopy, we found that T cells directly contact neurons in the hippocampus of Tau-P301S, 5xFAD, APP-KI, and CK-p25 mice. These T cell - neuron interactions were also observed in the prefrontal cortex and hippocampus of human AD brains. Using the CK-p25 model, which harbors neuronal DNA double-strand breaks (DSBs), we found that T cell infiltration coincides with periods of high neuronal DSBs. We also found that the majority of CD8 T cells contact neurons and are within less than 10 micrometers of a DSBs-bearing neuron. RNA sequencing of DSBs-bearing neurons from mice and humans revealed that MHC-I antigen presentation machinery is upregulated in these neurons. In CK-p25 mice, neuron-specific shRNA knock-down of B2M decreased CD8 T cell counts, neuronal volume, and synapsin expression in the dentate gyrus.
Furthermore, by bidirectionally manipulating neuronal activity, we demonstrated a functional link between T cells and neurons. We found that chronic neuronal activation with hM3Dq DREADDs in wild-type mice increased T cell infiltration into the brain. Conversely, dampening neuronal activity using the antiepileptic drug Levetiracetam decreased T cell infiltration in the Tau-P301S mice. Collectively, our findings suggest that T cell infiltration into the brain parenchyma and CD8 T cell - neuron interactions in AD are governed by neuronal DNA damage and neuronal hyperactivity.
Presenter: Amanda Fath
Advisor: Guoping Feng
Mutations of Autism-linked genes drive shared perceptual deficits across species
Abstract: Atypical sensory processing is a hallmark of autism spectrum disorder (ASD) and strongly predicts other core symptoms. However, the underlying computational and neural mechanisms remain poorly understood, in part due to the difficulty of modeling complex ASD symptoms in animals. Synchronized cross-species tasks provide a unique opportunity to quantitatively compare symptomatology in humans with autism and corresponding animal models. Here we use a recently developed cross-species sensory integration task across humans, mice, and marmosets to study perceptual deficits associated with Shank3 mutations, which underlie Phelan-McDermid syndrome (PMS). The human cohort spanned the autism spectrum, including individuals with PMS, and was complemented by parallel studies in Shank3 mutant and wild-type (WT) mice and marmosets. High-throughput behavior acquisition allowed for the generation of a large dataset (136 humans, 66 mice and 46 marmosets), and the task was playable across all species and groups, including humans with more severe ASD, such as PMS. Quantitative comparison indicated similar deficits in perception and learning, across humans, mice and marmosets with Shank3 mutations. Drift diffusion model (DDM) analysis revealed that these deficits are due to a decrease in drift rate, consistent with altered perceptual integration. In humans, gameplay statistics including drift rate, correlated with clinically relevant symptoms, assessed by traditional survey scores. Together these data indicate that Shank3 mutant mice and marmosets mimic specific perceptual deficits found in ASD and highlight the value of cross-species tasks for translational research and the discovery of disease mechanisms.