12023 research outputs found
Sort by
Mechanisms of Xist-Mediated Gene Silencing During the Initiation and Maintenance of X Chromosome Inactivation
Get PDFX chromosome inactivation (XCI) is a critical development process during which one of the two X chromosomes in female mammals is silenced to balance gene expression with males. XCI is initiated by upregulation of the long noncoding RNA (lncRNA) Xist from the future inactive X chromosome (Xi), which recruits a variety of proteins in cis to mediate transcriptional repression that is maintained throughout the lifetime of the organism. Recent studies have demonstrated that silencing following Xist expression is dependent on direct recruitment of the transcriptional silencing protein SHARP (also known as SPEN); however, the mechanism underlying formation of the Xi silencing compartment has remained poorly defined. Similarly, it has long been thought that maintenance of XCI occurs independently of Xist and depends on differential DNA methylation enrichment on the Xi, but the evidence in support of these views is lacking. Here, we show how low copy numbers of Xist can recruit SHARP in super-stoichiometric excess to initiate gene silencing on the X and mediate formation of the silent Xi compartment. We also provide preliminary evidence suggesting that maintenance of XCI is Xist independent, but dependent on DNA methylation and histone deacetylation. Together, these results offer a more holistic view of the molecular mechanisms underlying both initiation and maintenance XCI, as well as provide a framework for further investigation into lncRNA biology and epigenetic regulation more broadly
Defining the Universe of Functional RNA-Protein Interactions
Get PDFRNA has been proposed to mediate many central mechanisms of cell biology, including protein recruitment to chromatin, genome structure organization, and gene expression. In most cases, these critical functions have been widely attributed to the proteins to which RNAs bind. One paradigm example of this is the Xist long non-coding RNA, which complexes with many distinct proteins to orchestrate X-chromosome inactivation. Beyond Xist, there are many critical non-coding RNAs (ncRNAs) that are not yet functionally characterized because we lack information on what proteins they bind to. In this thesis, Chapter 1 discusses the growing gap between the vast potential of ncRNA functions and what has been demonstrated to be functionally meaningful. We highlight critical discrepancies between biochemical evidence supporting specific RNA-protein interactions and genetic evidence demonstrating the same interactions are often dispensable for function. Chapter 2 explores previously reported RNA-protein interactions for many chromatin proteins (i.e., PRC2, CTCF, etc.), demonstrating that they do not represent bona fide interactions in cells. We present Covalent Linkage Affinity Purification (CLAP), a method that employs denaturing purification of RNA-protein complexes, showing that CLAP accurately removes false signals that do not occur in vivo, while retaining known RNA-protein interactions. Chapter 3 details a highly multiplexed method of mapping RBPs and their in vivo binding sites across dozens to hundreds of targets within a single experiment. We present Split and Pool Identification of RBP targets (SPIDR), which enables the rapid, de novo discovery of RNA-protein interactions at an unprecedented scale and separates bona fide RBPs from non-RBPs. Using SPIDR, we uncover a previously unknown LARP1 binding site on the 18S ribosomal RNA that is directly adjacent to the mRNA entry channel, which may explain how LARP1 achieves translational control of sequence-specific mRNAs. Finally, Chapter 4 proposes new experimental and analytical approaches to evaluate the potentially wide universe of ncRNA-protein functions at scale. Together, these results provide a comprehensive framework for evaluating RNA-protein interactions and underscore the growing importance of RNA-mediated functions in cell biology
Essays in Empirical Industrial Organization
Get PDFThis dissertation comprises three essays related to the field of Empirical Industrial Organization. Chapter 1 and 2 contribute to the economic literature on online advertising auctions, and Chapter 3 contributes to the study of decision-making under risk using structural methods.
In Chapter 1, co-authored with Miguel Alcobendas, we provide a novel empirical analysis of a large-scale sequential market employing auctions to allocate objects to firms with budget constraints. Leveraging a unique proprietary dataset of online ad auctions, we examine the trade-off participants face due to short-run budget constraints. We develop and estimate a finite-horizon dynamic game among bidders with heterogeneous budgets, and we find that dynamic incentives significantly influence their participation and bidding strategies. We conduct a counterfactual simulation comparing first-price and second-price formats, illustrating how dynamics lead to significant disparities in competitive outcomes.
In Chapter 2, co-authored with Miguel Alcobendas, Matthew Shum, and Ke Shi, we investigate the impact of removing third-party cookies on the online advertising market. Utilizing a proprietary dataset of online ad auctions, we document stylized facts about the value of third-party cookies to advertisers. Adopting a structural approach, we simulate counterfactual scenarios to quantify the impact of Google's plan to phase out third-party cookies from Chrome. Our analysis suggests a 54\% reduction in publisher revenue and a 40\% reduction in advertiser surplus under an outright ban. Introduction of alternative tracking technologies under Google's Privacy Sandbox initiative would mitigate some of the loss. We find big tech firms can leverage their informational advantage to gain a larger surplus from the ban.
In Chapter 3, co-authored with Aldo Lucia, we explore the limited ability of prominent economic models in explaining multiple behavioral patterns. Conducting an experiment with 500 participants, we study two classical behaviors inconsistent with Expected Utility: the common ratio effect and preferences for randomization. We illustrate the lack of generalizability of existing models across these behaviors. Motivated by this, we introduce a novel empirical approach that does not commit on specific decision models. Our method offers more accurate out-of-sample predictions about behaviors under risk, both inside and outside laboratory settings, compared to leading economic models and machine learning algorithms.</p
Studies of the Evolution and Stability of the Thin Film Equation for Externally Modulated Control of Electrohydrodynamic and Thermocapillary Patterning
No full textIt has been known for a couple decades, based on extensive experimental, theoretical and numerical studies, that a flat slender nanoscale viscous film in the absence of gravity always undergoes early time linear instability when subject to electrical or thermocapillary forces. The patterns resulting from a uniform transverse electric or thermal field resemble clusters of small rounded protrusions whose early time dynamics have been described by linear stability analysis of the governing fourth-order nonlinear interface equation -- the so-called thin film equation. However, the pattern formation process beyond early times generates larger amplitude protrusions prone to coalescence or an Oswald-like ripening of adjacent formations which destroy the pattern uniformity. Introduction of film interface modulation by external spatially periodic modulation offers a superior method for this type of lithographic patterning. The resulting linear and nonlinear response of the liquid layer can be tuned to corral the evolution of the liquid interface into periodic arrays containing identical components in certain parameter range. Conditions for achieving high-fidelity patterns are still not fully understood, however, rendering such technique not yet fully utilized in practical applications.
To that end, we have conducted a number of analytical and numerical studies which elucidate various regimes leading to high-fidelity patterning by external spatial and temporal modulation. We focus on a single layer of viscous liquid film on a solid substrate which is described by the thin film equation derived under the long wavelength approximation. We first study the linear stability of periodic non-uniform stationary states subject to electrostatic stress and find that the necessary conditions for achieving stable states in 1D are the mass-limitation or saturation with a system-confining boundary (touching the mask) in order to suppress the coalescence and Ostwald-like-ripening modes. In 2D, stationary ridges are only achieved by saturation with a system-confining boundary in order to suppress its breakup. Time-dependent simulations further reveal inaccessible stationary states due to large electrode separation or large applied voltage. Exploratory studies on system subject to temperature gradient shows that the coalescence mode becomes unstable over a wider range of parameters due to thermocapillary stress. These findings result in phase diagrams relating the spatial modulation amplitude and electric Weber number or Marangoni number to the conditions for high-fidelity patterns which cannot be explained simply by matching the patterning and intrinsic instability wavelengths as previously claimed in literature.
We then turn to the optimal control of electrohydrodynamic thin film patterning where the optimal strategy in deforming a flat film toward a desired shape is determined. A computational framework is derived which allows us to study the open-loop terminal control problem for thin liquid film. The approach allows us to quantify the best-possible outcome only constrained by the underlying physical mechanisms, and better understand the limitations of thin film patterning in relation to the choice of target shapes and system parameters. The impact of imperfect engineering and methods of mitigations are also discussed, which should prove useful to soft lithography and other applications.</p
Quantum Error Correction Using Low-Density Parity-Check Codes and Erasure Qubits
Get PDFQuantum error correction is a method to reduce the effective error rate on quantum computers so that they can be used to carry out useful computation. In this thesis, we study two main problems: decoding quantum low-density parity-check codes and using erasure qubits to implement error correction protocols.
In the first part of this thesis, we focus on quantum low-density parity-check codes, which are a promising approach to reducing the spacetime overhead associated with error correction. We show that certain families of codes with constant rate and linear distance can be decoded efficiently. In particular, we propose a linear-time algorithm that will correct any error affecting at most a constant fraction of the qubits.
We also analyze the setting where the measurement outcomes given to the decoder can be corrupted. In this more realistic scenario, the decoder is shown to have the single-shot property. Using one round of noisy syndrome data, it can output a correction that is close to the data error as long as at most a constant fraction of the data qubits and syndrome bits are flipped. As a consequence, the decoder can operate under a stochastic noise model where errors occur with sufficiently small but constant probability.
In the second part of the thesis, we analyze quantum error-correcting codes implemented using erasure qubits. The idea behind erasure qubits is to bias the noise into a form where likely locations of errors are known, for example, by converting the dominant noise source into detectable leakage from the computational subspace. We provide a formalism for simulating and decoding stabilizer circuits with erasures, erasure checks, and resets. Using this formalism, we study the performance of Floquet codes and show that the benefits of knowing error locations outweigh the cost of extra noise due to erasure checks.
Lastly, we optimize erasure check schedules in the context of the surface code. By performing simulations with one, two, or four erasure checks per syndrome extraction round, we find different error parameter regimes where it is optimal to use each schedule. Additionally, we provide a simplified way of decoding erasure circuits suitable for circuits with infrequent erasure checks.</p
The Interplay of Waves and Stellar Evolution
Get PDFIn this thesis, I study the evolution of stellar interiors and stellar oscillations in order to address current observational puzzles in astronomy. The thesis focuses on refining physical models for pre-supernova outbursts from massive stars, tidal evolution of planetary architectures, and progenitors of compact neutron star binaries. In order to research these topics, I combine calculations of internal stellar oscillation modes with stellar evolution models, to account for the evolution of the modes across the stars' lives. I begin by introducing some concepts in stellar evolution and internal stellar oscillations which underlie the physical intuition and models present throughout the thesis.
In the second chapter of the thesis, I present a study of tidal dissipation in M-dwarfs hosting nearby exoplanets. I model dynamical tides from normal modes of stellar oscillation across stellar evolution. With my novel methods, I am able to resolve the detailed spectrum of tidal dissipation as a function of stellar age. This empowers our evolutionary calculations to capture the resonance locking phenomenon, in which sustained tidal excitation of modes to large amplitudes over long intervals of stellar evolution produces enhanced dissipation. I find that Earth-mass and Jupiter-mass planets around M-dwarfs experience significant orbital migration under the influence of resonance locking with inertial modes of the star.
In the third and fourth chapters of the thesis, I explore the ability of waves generated by vigorous core convection in massive stars to impart heat to the stellar envelopes. I model the excitation and propagation of waves during phases of energetic nuclear burning in stars to assess the amount of energy that waves transmit to the envelope, as well as the timescale before core collapse when the majority of wave heating occurs. I find that wave heating is unlikely to independently produce very massive circumstellar material (CSM), but induces large expansion that could trigger interaction with a binary companion and thereby drive the intense mass loss that is expected to precede interacting supernovae.
The fifth chapter explores this very mechanism of binary interaction to produce pre-supernova outbursts. Even without wave heating, stripped stars can expand greatly in the years before core collapse simply as a consequence of stellar evolution. I employ binary stellar evolution simulations to study how stripped stars of around 2-3 solar masses interact with neutron star companions, crucially focusing on the often-omitted stages from oxygen/neon (O/Ne) burning onward. I observe the stripped stars to undergo extremely high rates of mass loss, which can form a distribution of dense CSM around the system. My estimates for the CSM mass and radius are consistent with observed low ejecta-mass, interacting supernovae.</p
Composition, Structure, and Formation of the Lower Crust in Continental and Oceanic Arc Settings: Insights from the Xenolith Record
Get PDFThe compositional variability of lavas erupted in subduction zone settings results from a multitude of deep crustal processes acting in concert, reflecting variations among arcs in source rock composition, water content, oxidation state (fO2), temperature, pressure, and crystallization sequence across arcs. Lower to mid-crustal xenoliths—deeply sourced rock fragments entrained and brought to the surface by ascending melts—provide a robust record of the chemistry and structure of their inaccessible source regions. This thesis combines a variety of laboratory techniques and modeling approaches to explore the origins of two xenolith suites from vastly different arc settings: the oceanic Aleutian Arc off the coast of Alaska and the continental Andean Arc in Colombia. While all samples are fully characterized in terms of their petrography and the major and trace element chemistry (of both minerals and whole-rock), special attention is given to stable Fe isotope ratios, which are particularly sensitive to the fractionation of important Fe-bearing minerals like olivine, magnetite, amphibole, and garnet.
In the first chapter, we document the major and trace element compositions of 39 previously undescribed xenoliths from the Mt. Moffett and Mt. Adagdak volcanic centers on Adak Island, Central Aleutians. This data is then used to evaluate the P-T-fO2-H2O conditions under which the cumulates formed and interrogate the nature of their parental melts. The second chapter builds upon this first study, presenting Fe isotope data of whole-rock powders and mineral separates (spinel, clinopyroxene, olivine, amphibole, and magnetite) from the Adagdak xenoliths. Our data show that the Adak crust is stratified in terms of Fe isotopes, with an isotopically light lower-crust and an isotopically heavy middle to upper-crust. The implications of this compositional structure and its relation to the evolution of Adagdak magmas is then explored through mass-balance fractional crystallization modeling. In the final chapter, we apply the same methods used in the first two chapters to characterize a suite of lower to mid-crustal xenoliths from the Mercaderes region of Colombia in the Central Andean Cordillera. In contrast to Adagdak, the Mercaderes samples show a nearly constant whole-rock Fe isotope composition throughout our ~50 km crustal section. Through thermodynamic modeling, we show that the most likely explanation for this data is that the Mercaderes suite represents a prograde metamorphic sequenc
ChIP-DIP: a Multiplexed Method for Mapping Proteins to DNA Uncovers Combinatorics Controlling Gene Expression
Get PDFGene regulation is governed by the complex interplay between thousands of regulatory proteins and chromatin states; understanding how these dynamics give rise to precisely controlled, cell type-specific gene expression has been a central goal of molecular biology. Yet, addressing this goal remains challenging because current methods for mapping proteins to DNA are labor-intensive, resource-demanding, and limited to studying a single or a small number of proteins at a time. To overcome this, we developed ChIP-DIP (ChIP Done In Parallel), a novel split-pool-based method that enables simultaneous, genome-wide mapping of hundreds of diverse regulatory proteins in a single experiment. We demonstrate that ChIP-DIP generates highly accurate maps equivalent to traditional approaches, with data quality unaffected by the number of distinct proteins or the composition of proteins measured within a single experiment. We show that, because of this multiplexed capability, ChIP-DIP enables generation of highly accurate maps using several orders of magnitude fewer cells per protein compared to traditional approaches (~30,000 fold), making it a powerful tool for studying a diverse array of proteins–DNA interactions with limited cellular input. In addition, we show that ChIP-DIP can generate high-quality maps for all classes of DNA-associated proteins, including histone modifications, chromatin regulators, transcription factors, and RNA polymerases. Using these data, we explore quantitative combinations of histone modifications and integrate these signatures with RNA polymerase activity, chromatin regulatory protein binding and transcription factor binding to define distinct classes of regulatory elements (e.g., distinct types of enhancer elements), their functional activity (e.g., transcriptional activity), and their regulatory potential (e.g., poised for activation upon stimulation or differentiation). Together, our results demonstratethat ChIP-DIP enables generation of consortium-level data within a single lab and highlight the importance of this approach for studying mechanisms of gene regulation in a context and cell type-specific manner. Lastly, ChIP-DIP provides a powerful platform to multiplex protein detection and provides a unique opportunity to incorporate other split-pool-based assays such as SPRITE and single-cell SPRITE to detect protein specific nuclear structures and multiplexed single-cell chromatin profiles, respectively. This work represents a transformational framework on how to study biology in a holistic manner
Structural and Dynamical Correlations Linked to Smaller Thermal Resistance at a Classical Liquid/Solid Interface
No full textEver more powerful and densely packed chips for applications like cryptocurrency mining and artificial intelligence generate such enormous heat fluxes that designers are pivoting from gas to liquid cooling to forestall damage from thermal runaway. Even with optimal flow patterns however, the intrinsic thermal boundary resistance at the liquid/solid (L/S) interface poses an additional source of thermal impedance. There is a lingering misconception in the field that the higher the liquid contact density, the more frequent the L/S collision rate and the smaller the thermal slip length. Here we present an insightful counterexample based on non-equilibrium molecular dynamics simulations of a classical liquid confined between different facets of a face centered cubic crystal held at different temperature. We have conducted a comprehensive study to quantify thermal exchange and propagation across the interface by varying the L/S interaction energy, L/S repulsive distance, facet orientation, thermal flux and local temperature with particular emphasis on the properties of the liquid contact layer (i.e., liquid monolayer adjacent to the solid surface). Numerous static and dynamic quantities characterizing the contact layer reveal the ways in which long range order, anisotropy of the L/S potential and correlated motion act to reduce the thermal slip length. Systems with the smallest thermal slip length exhibit two distinct features: 2D caged motion with string-like alignment of liquid particles unlike that observed in glassy systems and larger non-ergodicity parameter but shorter, not longer, caging times.
These simulations have revealed two master curves which help unify the various influences at play. The first relation directly links the thermal slip length to the temperature modified 2D static structure factor representing long-range order in the contact layer. The second relation directly links the thermal slip length to the temperature modified dominant frequencies of the first solid and liquid layer as extracted from the density of states. These correlations, which represent power law dependencies, offer a new paradigm for the design of L/S interfaces to maximize thermal exchange across a classical L/S interface.</p
Learning-Based Perception for Robotics in Suboptimal Data Landscapes
Get PDFAutonomous robots are increasingly present in the world today, being used across a variety of settings and applications. In order to interact with their surroundings, robots typically use cameras to see the world, employing computer vision algorithms to comprehend rich, visual information. While contemporary, learning-based computer vision models provide robots with an accurate and robust understanding of their surroundings, most off-the-shelf methods rely on supervised deep learning techniques, requiring abundant labeled data in order to train and prevent overfitting. However, in many robotic applications and settings, the data landscape is characterized by data scarcity and/or the lack of apparent supervisory signals. Since custom perception solutions are often required for robotic applications, direct adoption of common computer vision methods proves challenging.
In this thesis, we develop robotic perception approaches across three different applications that overcome the challenges of such data landscapes. First, we develop learning-based visual terrain-relative navigation (VTRN) approaches for high-altitude aerial vehicles. This is a problem for which relevant data is available, but made difficult by the lack of obvious supervisory signals related to the high-level navigation objective. In the first chapters of the thesis, we show the power of self-supervised learning approaches to increase VTRN robustness to seasonal and temporal variations that would otherwise debilitate such systems.
Next, we address the challenge of developing thermal semantic perception algorithms for aerial field robotics. Due to the specialized nature of field environments and the sensing modality, development of thermal vision algorithms under these conditions is often characterized by the lack of relevant data. We show how we develop various thermal semantic segmentation in response to the evolving data constraints inherent in field robotic projects. In the final part of the thesis, we develop data-efficient, multispectral deep learning algorithms for autonomous driving applications where the lack of data arises from the need for custom, multispectral datasets that are synchronized and coregistered.</p