Publications

Registration of histology images from multiple sources is a pressing problem in large-scale studies of spatial -omics data. Researchers often perform “common coordinate registration,” akin to segmentation, in which samples are partitioned based on tissue type to allow for quantitative comparison of similar regions across samples. Accuracy in such registration requires both high image resolution and global awareness, which mark a difficult balancing act for contemporary deep learning architectures. We present a novel convolutional neural network (CNN) architecture that combines (1) a local classification CNN that extracts features from image patches sampled sparsely across the tissue surface, and (2) a global segmentation CNN that operates on these extracted features. This hybrid network can be trained in an end-to-end manner, and we demonstrate its relative merits over competing approaches on a reference histology dataset as well as two published spatial transcriptomics datasets. We believe that this paradigm will greatly enhance our ability to process spatial -omics data, and has general purpose applications for the processing of high-resolution histology images on commercially available GPUs.

Most cognitive processes are studied using abstract or synthetic stimuli with specific features to fully control what is presented to subjects. However, recent studies have revealed enhancements of cognitive capacities (such as working memory) when processing naturalistic versus abstract stimuli. Using abstract stimuli constructed from distinct visual features (e.g., color and shape), we have recently shown that human subjects can learn multidimensional stimulus-reward associations via initially estimating reward value of individual features (feature-based learning) before gradually switching to learning about reward value of individual stimuli (object-based learning). Here, we examined whether similar strategies are adopted during learning about naturalistic stimuli that are clearly perceived as objects (instead of a combination of features) and contain both task-relevant and irrelevant features. We found that similar to learning about abstract stimuli, subjects initially adopted feature-based learning more strongly before transitioning to object-based learning. However, there were three key differences between learning about naturalistic and abstract stimuli. First, compared with abstract stimuli, the initial learning strategy was less feature-based for naturalistic stimuli. Second, subjects transitioned to object-based learning faster for naturalistic stimuli. Third, unexpectedly, subjects were more likely to adopt feature-based learning for naturalistic stimuli, both at the steady state and overall. These results suggest that despite the stronger tendency to perceive naturalistic stimuli as objects, which leads to greater likelihood of using object-based learning as the initial strategy and a faster transition to object-based learning, the influence of individual features on learning is stronger for these stimuli such that ultimately the object-based strategy is adopted less. Overall, our findings suggest that feature-based learning is a general initial strategy for learning about reward value of all types of multi-dimensional stimuli.

Living matter moves, deforms, and organizes itself. In cells this is made possible by networks of polymer filaments and crosslinking molecules that connect filaments to each other and that act as motors to do mechanical work on the network. For the case of highly cross-linked filament networks, we discuss how the material properties of assemblies emerge from the forces exerted by microscopic agents. First, we introduce a phenomenological model that characterizes the forces that crosslink populations exert between filaments. Second, we derive a theory that predicts the material properties of highly crosslinked filament networks, given the crosslinks present. Third, we discuss which properties of crosslinks set the material properties and behavior of highly crosslinked cytoskeletal networks. The work presented here, will enable the better understanding of cytoskeletal mechanics and its molecular underpinnings. This theory is also a first step towards a theory of how molecular perturbations impact cytoskeletal organization, and provides a framework for designing cytoskeletal networks with desirable properties in the lab.

Multitaper estimators have enjoyed significant success in estimating spectral densities from finite samples using as tapers Slepian functions defined on the acquisition domain. Unfortunately, the numerical calculation of these Slepian tapers is only tractable for certain symmetric domains, such as rectangles or disks. In addition, no performance bounds are currently available for the mean squared error of the spectral density estimate. This situation is inadequate for applications such as cryo-electron microscopy, where noise models must be estimated from irregular domains with small sample sizes. We show that the multitaper estimator only depends on the linear space spanned by the tapers. As a result, Slepian tapers may be replaced by proxy tapers spanning the same subspace (validating the common practice of using partially converged solutions to the Slepian eigenproblem as tapers). These proxies may consequently be calculated using standard numerical algorithms for block diagonalization. We also prove a set of performance bounds for multitaper estimators on arbitrary domains. The method is demonstrated on synthetic and experimental datasets from cryo-electron microscopy, where it reduces mean squared error by a factor of two or more compared to traditional methods.

A major obstacle to treating Alzheimer’s disease (AD) is our lack of understanding of the molecular mechanisms underlying selective neuronal vulnerability, a key characteristic of the disease. Here, we present a framework integrating high-quality neuron-type-specific molecular profiles across the lifetime of the healthy mouse, which we generated using bacTRAP, with postmortem human functional genomics and quantitative genetics data. We demonstrate human-mouse conservation of cellular taxonomy at the molecular level for neurons vulnerable and resistant in AD, identify specific genes and pathways associated with AD neuropathology, and pinpoint a specific functional gene module underlying selective vulnerability, enriched in processes associated with axonal remodeling, and affected by amyloid accumulation and aging. We have made all cell-type-specific profiles and functional networks available at http://alz.princeton.edu. Overall, our study provides a molecular framework for understanding the complex interplay between Aβ, aging, and neurodegeneration within the most vulnerable neurons in AD.

A generic flow distribution network typically does not deliver its load at a uniform rate across a service area, instead oversupplying regions near the nutrient source while leaving downstream regions undersupplied. In this work we demonstrate how a local adaptive rule coupling tissue growth with nutrient density results in a flow network that self-organizes to deliver nutrients uniformly. This geometric adaptive rule can be generalized and imported to mechanics-based adaptive models to address the effects spatial gradients in nutrients or growth factors in tissues.

Hot spots, or plasmoids, which form due to magnetic reconnection in current sheets, are conjectured to power frequent X-ray and near-infrared flares from Sgr A*, the black hole in the center of our Galaxy. It is unclear how, where, and when current sheets form in black hole accretion disks. We perform axisymmetric general-relativistic resistive magnetohydrodynamics simulations to model reconnection and plasmoid formation in a range of accretion flows. Current sheets and plasmoids are ubiquitous features that form regardless of the initial magnetic field in the disk, the magnetization in the quasisteady-state phase of accretion, and the spin of the black hole. Within 10 Schwarzschild radii from the event horizon, we observe plasmoids forming, after which they can merge, grow to macroscopic scales of the order of a few Schwarzschild radii, and are ultimately advected along the jet’s sheath or into the disk. Large plasmoids are energized to relativistic temperatures via reconnection and contribute to the jet’s limb brightening. We find that only hot spots forming in magnetically arrested disks can potentially explain the energetics of Sgr A* flares. The flare period is determined by the reconnection rate, which we find to be between $0.01c$ and $0.03c$ in all cases, consistent with studies of reconnection in isolated Harris-type current sheets. We quantify magnetic dissipation and nonideal electric fields, which can efficiently inject nonthermal particles. We show that explicit resistivity allows for converged numerical solutions, such that the electromagnetic energy evolution and dissipation become independent of the grid scale for the extreme resolutions considered here.

We introduce a procedure to systematically search for a local unitary transformation that maps a wave function with a nontrivial sign structure into a positive-real form. The transformation is parametrized as a quantum circuit compiled into a set of one- and two-qubit gates. We design a cost function that maximizes the average sign of the output state and removes its complex phases. The optimization of the gates is performed through automatic differentiation algorithms, widely used in the machine learning community. We provide numerical evidence for significant improvements in the average sign for a two-leg triangular Heisenberg ladder with next-to-nearest-neighbor and ring-exchange interactions. This model exhibits phases where the sign structure can be removed by simple local one-qubit unitaries, but also an exotic Bose-metal phase whose sign structure induces “Bose surfaces” with a fermionic character and a higher entanglement that requires deeper circuits.

Loss of normal tissue architecture is a hallmark of oncogenic transformation1. In developing organisms, tissues architectures are sculpted by mechanical forces during morphogenesis2. However, the origins and consequences of tissue architecture during tumorigenesis remain elusive. In skin, premalignant basal cell carcinomas form ‘buds’, while invasive squamous cell carcinomas initiate as ‘folds’. Here, using computational modelling, genetic manipulations and biophysical measurements, we identify the biophysical underpinnings and biological consequences of these tumour architectures. Cell proliferation and actomyosin contractility dominate tissue architectures in monolayer, but not multilayer, epithelia. In stratified epidermis, meanwhile, softening and enhanced remodelling of the basement membrane promote tumour budding, while stiffening of the basement membrane promotes folding. Additional key forces stem from the stratification and differentiation of progenitor cells. Tumour-specific suprabasal stiffness gradients are generated as oncogenic lesions progress towards malignancy, which we computationally predict will alter extensile tensions on the tumour basement membrane. The pathophysiologic ramifications of this prediction are profound. Genetically decreasing the stiffness of basement membranes increases membrane tensions in silico and potentiates the progression of invasive squamous cell carcinomas in vivo. Our findings suggest that mechanical forces—exerted from above and below progenitors of multilayered epithelia—function to shape premalignant tumour architectures and influence tumour progression.

In the present paper we describe a class of algorithms for the solution of Laplace's equation on polygonal domains with Neumann boundary conditions. It is well known that in such cases the solutions have singularities near the corners which poses a challenge for many existing methods. If the boundary data is smooth on each edge of the polygon, then in the vicinity of each corner the solution to the corresponding boundary integral equation has an expansion in terms of certain (analytically available) singular powers. Using the known behavior of the solution, universal discretizations have been constructed for the solution of the Dirichlet problem. However, the leading order behavior of solutions to the Neumann problem is $O(t^{\mu})$ for $\mu \in (-1/2,0)$ depending on the angle at the corner (compared to $O(C+t^{\mu})$ with $\mu>1/2$ for the Dirichlet problem); this presents a significant challenge in the design of universal discretizations. Our approach is based on using the discretization for the Dirichlet problem in order to compute a solution in the "weak sense" by solving an adjoint linear system; namely, it can be used to compute inner products with smooth functions accurately, but it cannot be interpolated. Furthermore we present a procedure to obtain accurate solutions arbitrarily close to the corner, by solving a sequence of small local subproblems in the vicinity of that corner. The results are illustrated with several numerical examples.