“The multi-scale brain: modeling the impact of local structure and dynamics on brain function and dysfunction”

Abstract: The brain’s ability to carry out metabolically-intensive complex computations, at the level of neural circuits, is supported by its local cellular properties.  Local neuronal properties, such as synaptic organization, neuronal morphology, and neurotransmitter dynamics, have a significant impact on neuroplasticity, metabolism, and neuronal health. These local effects translate to upstream effects on brain-wide function, influencing learning and development, and even contributing to the initiation and progression of neurodegenerative diseases. However, there is a notable lack of modeling work, that can abstract and integrate these scales to reveal the principles of neurobiologically-constrained brain function. My research seeks to fill this gap by developing multi-scale models that connect cellular mechanisms with brain-wide function, to provide the basis of biologically-plausible and metabolically-efficient brain function.

In my talk, I will be covering two chief research directions from my work. First, I will be talking about how the morphology of single-neurons, in particular the arborization properties of its dendritic trees, could impact the ‘computational complexity’ of the calculations that a neuron can perform locally. I will present a novel deep network formalism to link the dendritic architecture across diverse cell types, to the neuronal cell type’s computational capabilities. I used this model of a `dendritic circuit’ to show that a single neuron can locally compute complex functions in very high input dimensions. Intriguingly, the organization of dendritic compartments in a size-constrained circuit can lead to qualitatively different computational outcomes, thereby mapping the morphological properties of a neuronal cell type to its computational specialization.

While the ability of dendritic trees to perform non-linearly separable computations has been demonstrated earlier, my work is the first one to offer several order parameters to quantify and characterize the computational complexity of a morphological cell type along the axes of effective input dimensionality, and the entropy and sensitivity of the computational function that the cell is capable of performing.

Secondly, I will be talking about my efforts to integrate the biophysical priors of neuropathology accumulation and propagation into a statistical inference framework to estimate the pseudotemporal disease trajectory of Alzhiemer’s Disease. I have been working with the Seattle Alzheimer’s Disease Consortium (SEA-AD) to develop novel methods incorporating multimodal neuropathological data, including multiple proteomic factors (such as Amyloid-Beta and tau polymeric aggregates) and cellular factors (such as local neuronal and non-neuronal densities and cellular colocalization with pathological proteins), measured across multiple brain regions.  Our findings have successfully defined a pseudotemporal trajectory of Alzheimer’s Disease in the brain’s MTG region, and will be applied to multiple brain regions in upcoming studies.

In conclusion, my research has been guided by a desire to connect small-scale neurobiological processes with larger brain functions. By employing multi-scale modeling techniques and collaborating across disciplines, I have been able to tackle questions that span from cellular dynamics to brain-wide health. Looking forward, I am excited to continue building on these foundational insights through novel quantitative and theoretical method development.

Bio: Anamika Agrawal joined the Allen Institute and the University of Washington as a Shanahan Foundation Fellow in August 2022. She received her Ph.D. in Physics from UC San Diego, working with Prof. Elena Koslover. During her graduate studies, she became interested in the interplay of intracellular dynamics and neuronal morphology, with her work focusing on mitochondrial organization. Drawing inspiration from her specialization in Quantitative Biology at UCSD, she enjoys working with simple analytical and computational models that still preserve the relevant biological complexity of the involved data. During the fellowship, she is interested in understanding how cell-type-specific neuronal properties, such as morphology and transcriptomics, contribute to neuronal function and dysfunction.

The 2024-2025 seminars will be held in person, and are free and open to the public.