71446 research outputs found
Sort by
Creating fiscal space for health in 30 low- and middle-income countries
Get PDFThis policy brief was developed as part of the collaboration between the Harvard Health Systems Innovation Lab and the Changing Diabetes in Children (CDiC) program, in collaboration with and supported by Novo Nordisk. The analysis currently focuses on the 30 countries where CDiC is active. The brief presents results on opportunities to expand fiscal space for health. A technical report and manuscript with detailed methodology and findings are forthcoming.Version of Recor
The Best of Three Worlds? Privacy, Welfare, and Fairness for Facility Location Mechanism Design
No full textThe facility location problem in economics examines where a public facility should be placed to maximize the total welfare of its prospective users. While this objective, termed social welfare, is the primary focus of traditional literature on facility location, other factors must also be considered in real-world implementations. One such desideratum is privacy: the welfare-maximizing location for the facility may be determined by highly personalized data about the population (e.g., where people live and which individuals are likely to utilize the facility), so it is important to ensure that the process (or "mechanism") of selecting the facility location does not inadvertently leak this sensitive information to the public.
However, to protect against such privacy risks, a facility location mechanism must reduce its reliance on individualized data. Consequently, the resulting privacy-preserving output will yield lower social welfare than the optimal spot for the facility, producing a welfare loss that can be viewed as the "cost of privacy." This illustrates a tradeoff between social welfare and privacy that existing research already covers. Yet, the imposition of privacy also induces a third consideration that has not been similarly studied: fairness in how the "cost of privacy" is distributed across individuals. For instance, suppose we can make a mechanism private while only incurring a small loss in social welfare, but this welfare loss is entirely borne by a single individual. Even though this implementation minimizes the privacy-social welfare tradeoff, it would still be undesirable as it disproportionately harms select members of the population. In this thesis, we therefore consider three objectives simultaneously and ask: What is the tradeoff between privacy, social welfare, and fairness when designing mechanisms for facility location?
We quantify privacy through the framework of differential privacy, utilize a standard measure of social welfare, and propose a novel measure of fairness. Under this setup, we first derive an impossibility result that privacy and fairness cannot be simultaneously guaranteed over all possible datasets that could represent the locations of individuals in a population. We then offer a relaxation of the original problem that only seeks fairness and social welfare over smaller, more "realistic-looking" families of datasets. For this relaxation, we construct a private mechanism M and prove high probability upper bounds on its loss with respect to fairness and social welfare. At the same time, we derive information-theoretic lower bounds on the amount of fairness and social welfare loss that any differentially private mechanism must incur. A comparison of these bounds shows that in addition to being differentially private, our mechanism M is simultaneously optimal (or, for a harder family of datasets, near-optimal up to small factors) on fairness and social welfare. This suggests that while there is a tradeoff between privacy and each of social welfare and fairness, there is no additional tradeoff when we consider all three objectives simultaneously, provided that the population data is sufficiently "natural."Computer Scienc
Using Novel Genetic Methods to Better Understand the Associations of Behavior and Brain with Psychiatric Phenotypes
No full textPsychiatric disorders comprise a heterogeneous group of conditions that profoundly affect long-term health, cognitive functioning, and overall well-being. Large-scale genome-wide association studies (GWAS) have demonstrated that these disorders have a highly polygenic genetic architecture. At the same time, increasingly powerful GWAS and whole-exome sequencing studies of related phenotypes, including lifestyle factors and brain morphology, provide new opportunities to elucidate the biological pathways linking genetics, brain structure, behavior, and psychopathology. This dissertation leverages advances in GWAS and exome sequencing to investigate the genetic architecture underlying psychiatric symptoms and their correlates across behavioral and neuroanatomical domains, with the aim of clarifying mechanistic links and identifying potential therapeutic targets.
In Chapter 1, we evaluated genetic confounding (i.e., genetic factors acting as a common cause) in the associations of sleep duration and physical activity with internalizing problems in adolescents. Using well-powered GWAS and genetically informed methods, the analysis distinguished direct behavioral effects from shared polygenic influences. The findings indicated substantial shared genetic influences between sleep duration and internalizing problems, which may in part reflect reporting-related measurement error arising from shared method variance rather than direct causal effects. In contrast, the associations between physical activity and internalizing problems were not genetically confounded, suggesting a more direct influence of this modifiable behavior after accounting for shared genetic liability.
In Chapter 2, we explored if genetic loci shared between major psychiatric disorders and brain morphology show regional or global effects. By integrating GWAS of 180 regions of cortical structures and six major psychiatric disorders, we identified substantial pairwise genetic overlaps of overall cortical surface area or thickness with psychiatric disorders, but no clear
directional pattern. Most genomic loci shared across all six disorders, showed opposing directional effects in different regions across the cortex while one locus showed a regional specific effect on reduced primary visual and posterior cingulate surface area. The directional heterogeneity showed the complex link between brain morphology and psychiatric disorders.
In Chapter 3, we investigated whole exome-wide associations of comprehensively characterized sleep phenotypes. These phenotypes include self-reported questionnaires, diagnoses and medication prescription extracted from electronic health records, and accelerometer-based measures. These rare-variant associations highlighted potential distinct biological pathways linked to different types of sleep phenotypes and suggested novel therapeutic avenues.
Collectively, studies integrate GWAS, exome sequencing, and deep phenotyping to explore how genetic variation contributes to psychiatric disorders. By examining the genetic underpinnings of behavior, brain structure, and psychopathology, this dissertation seeks to elucidate shared and distinct mechanisms across psychiatric conditions. By jointly leveraging causal inference, cross-disorder analyses, and deep phenotyping approaches, this work informs strategies for prevention and intervention in precision psychiatry.Population Health Science
Visualization and Interpretability for Multiplexed Spatial Biology
No full textSpatial biology investigates how cells organize into tissues and how that organization shapes function, disease progression, and response to therapy. Multiplexed tissue imaging makes these relationships visible by measuring dozens of markers in situ across whole-slide specimens, yielding gigapixel images with millions of cells. Analyzing these data at cohort scale is challenging: images are high dimensional and heterogeneous, and current workflows often separate computational results from the tissue context that gives them meaning.
Extracting insight at this scale requires visualization, interpretability, and analysis methods that keep multivariate measurements and AI model outputs grounded in tissue context and enable reproducible cohort-scale comparisons.
This dissertation develops methods that make multiplexed spatial imaging data more interpretable, reproducible, and scalable. We introduce psudo, a method and tool for principled pseudocoloring and palette assignment that optimizes color choices under perceptual and spatial criteria to produce faithful, task-appropriate composites. We present Visinity, a visual analytics system for cohort-scale neighborhood analysis that quantifies cell–cell interactions and supports exploratory and confirmatory analysis of recurrent interaction patterns. We contribute SEAL, an embedding interpretability approach that links 2D projections to their spatial images and underlying feature values, enabling users to interpret clusters and trace selections back to tissue context. Across studies with domain experts on human tonsil, mouse lung, and colorectal cancer data, these systems accelerated hypothesis generation and testing and improved communication of results. Together, they shorten the path from measurement to insight and expand the scope of questions that spatial biology can address.Engineering and Applied Sciences - Computer Scienc
Brainstem sensing of multiple body signals during food consumption
No full textStudies of body-to-brain communication often examine one stimulus or organ at a time, yet the brain must integrate many body signals during behavior. For example, food consumption generates diverse oral and post-oral chemical and mechanical signals transduced by well-characterized peripheral neuronal pathways. Far less is known about how these and other bodily signals are integrated and organized in the brainstem lateral parabrachial nucleus (LPBN), a key interoceptive sensory hub. We established methods to image the activity of 1000s of neurons throughout a large region of mouse LPBN. Food consumption drove a seconds-long wave of activity across LPBN, with dynamics mirroring the movement of food through the upper gastrointestinal tract observed using X-ray fluoroscopy. By imaging the same neurons across days, we found that spatially clustered subsets of neurons encoded oral signals, stomach filling, visceral malaise, arousal, and/or body movement. Moreover, only certain subsets were modulated by cortical input. Together, these experiments reveal a functional specialization in the LPBN that integrates contextual information from the body to guide behavior.Neuroscienc
Evolution of discrete reproductive strategies in Phlox (Polemoniaceae)
No full textFitness is defined as an individual’s relative success in reproduction and survival. As populations adapt to changing ecological and environmental conditions, the fitness landscape can favor divergent reproductive strategies, leading to correlated and cascading effects on reproductive biology. In flowering plants, maternal fitness depends on both seed size—a proxy for offspring quality—and seed number, reflecting offspring quantity. However, a fundamental tradeoff between seed size and number constrains the potential for simultaneous fitness gains in both traits. In this thesis, I use an eco-evo-devo framework to investigate how and why variation in seed number arises within the wildflower genus Phlox. In Chapter 1, I characterize the developmental basis of ovule number variation across five independent origins of a novel multi-ovulate ovule packaging morphology from a single-ovulate ancestral state. We show that a dramatic shift at the earliest stages of development in the allometric relationship between ovule size and ovary size governs the transition from a single-ovulate to a multi-ovulate state. Specifically, this allometric difference reflects two distinct developmental trajectories: one in which a larger ovary yields a larger, single seed and the other in which a larger ovary yields an increased number of smaller seeds. The convergent evolution of this novel developmental response in every instance of the derivation of the multi-ovulate condition may reflect a previously unappreciated constraint on ovule development. Furthermore, this single mechanism appears to be the basis of variation in the size and number of both seeds and ovules. In Chapter 2, I created an F2 hybrid population using crosses between sister species Phlox drummondii and P. roemeriana, representing the extremes of not only ovule packaging strategies but also correlated variation in flower size and number. Most measured traits in the F2 population show unimodal distributions, consistent with a genetic architecture involving multiple loci. Significant correlations between ovary size, ovule number, flower size, and flower number among F2 plants suggest that underlying genetic constraints may shape both the seed size/number and flower size/number tradeoffs, as well as the relationship between them. Finally, in Chapter 3, I characterize a natural polymorphism in Phlox longifolia involving the ancestral single-ovulate strategy and the derived multi-ovulate strategy. I show that environmental variation alone cannot explain the phenotypic differences observed across 102 populations spanning the species' range, suggesting a role of genetic variation. Collectively, these studies reveal the developmental underpinnings and the role of genetic variation—including integration with other phenotypic traits—of seed packaging strategies in Phlox, offering novel insight into how changes in fundamental developmental pathways generate morphological diversity in natural populations and influence key components of the tradeoff between seed size and number.Biology, Organismic and Evolutionar
Neuronal Control of Host Defense Against Bee Phospholipase A2
No full textNeuroimmune interactions have been increasingly recognized as key mechanisms in both promotion and modulation of mammalian host defense against external insults. In particular, skin-innervating TRPV1+ sensory neurons are known to directly detect harmful stimuli, including microbial products, toxins, and allergens, and respond by releasing neuropeptides, which signal to local innate immune cells to inform subsequent inflammatory, wound-healing, and/or adaptive immune responses. In this dissertation, I explore roles of sensory neurons in the allergic sensitization to bee venom allergen phospholipase A2 (bvPLA2) and in the innate host defense against its systemic toxicity.
Using footpad injections in wildtype mice, I confirm that bvPLA2 induces a type-2 allergic immune response, mediated by CD301b+ skin dendritic cells (DCs) and characterized by development of Th2 cells in the draining lymph node (dLN) and production of IgE. We also report that bvPLA2 directly activates dorsal root ganglia neurons, evidenced by release of the neuropeptide CGRP both in vitro and in vivo. Utilizing two methods of in vivo TRPV1+ neuron ablation, I show that TRPV1+ neurons dampen the migration of activated CD301b+ DCs in response to subcutaneous bvPLA2 immunization via a CGRP-independent mechanism and limit the subsequent development of Th2 cells in the dLN.
Besides its allergenicity, bvPLA2 is also highly toxic due to its ability to disrupt cell membranes and contributes to multi-organ damage during mass envenomation events. Here I also describe a crucial protective role of TRPV1+ neurons against rapid and systemic toxic effects of bvPLA2, wherein ablation of TRPV1+ neurons renders mice highly susceptible to death within 24 hours after a single subcutaneous injection of bvPLA2 at a normally sublethal dose. I also characterize the early local innate immune response to bvPLA2 in the skin and investigate a possible link between TRPV1+ neurons and bvPLA2-induced eosinophil influx and degranulation.
In summary, this thesis reports two distinct mechanisms through which TRPV1+ neurons control host defense responses to bvPLA2 as a venom toxin and allergen.Immunolog
A Biopsychosocial Model of Adolescent Gender Dysphoria
No full textGender dysphoria is a mental health condition characterized by clinically significant distress or impairment associated with marked gender incongruence, a feeling of discordance between one’s gender identity and biological sex. Gender incongruence and dysphoria prevalence have risen substantially in recent years, particularly in adolescent females. My dissertation aims to understand the etiology of these phenomena through a historical overview of clinical gender dysphoria, case studies of gender reassignment in populations with disorders of sex development, theories of the neurobiology of sexual development, and four empirical studies examining gender incongruence across early adolescent development. My empirical chapters make use of the Adolescent Brain Cognitive Development (ABCD) Study sample, a demographically diverse longitudinal sample of 11,864 youths in the United States followed from ages 9-10 years in 2016-2018 to ages 14-15 years in 2021-2023.
Chapter 1 reports developmental trajectories of self- and parent-reported gender incongruence, gender expression, and gender identity across five timepoints. While gender expression was stable with age, gender incongruence and transgender or nonbinary (TNB) self-identification increased with age, particularly in females. At baseline, 0.5% of males and 1.1% of females identified as TNB. By Wave 5, 1.2% of males and 9.6% of females identified as TNB. Females were also more likely to report gender incongruence and elevated gender nonconformity (i.e., sex-atypical gender expression). TNB youth reported elevated gender nonconformity across all ages, but showed increased gender incongruence only at later timepoints. Parents reported lower gender incongruence across all measures.
Chapter 2 makes use of ABCD’s large sample of 1,970 same-sex twins to estimate the heritability of gender incongruence. Gender incongruence was more common in females, and concordance was higher in female twin pairs than male twin pairs. Binary probandwise concordance for TNB gender identity was 46% among monozygotic pairs compared to 13% among dizygotic pairs. ACE models revealed that genetic factors explained 47% of the variance of gender incongruence, with the remaining variance explained by non-shared environmental factors. Estimates had large standard errors, and thus high uncertainty, but suggest a large genetic contribution to gender incongruence in adolescents.
Chapter 3 examined concurrent and longitudinal cross-lagged associations between self-reported gender incongruence and a range of biopsychosocial variables known or hypothesized to be associated with gender dysphoria: internalizing symptoms, autism spectrum traits, sexual orientation, pubertal timing, body mass index, family conflict, peer victimization, cyberbullying, screen time, and number of TNB friends. All predictors of interest were significantly associated with gender incongruence except family conflict and number of TNB friends. Cross-lagged models tested competing hypotheses between the theory of rapid-onset gender dysphoria (ROGD) and minority stress theory on the directionality of these associations. Higher internalizing symptoms and screen time predicted later gender incongruence, but not vice versa, in females across all timepoints, while other associations were mixed. Longitudinal results largely favored predictions made by ROGD over minority stress theory.
Chapter 4 aimed to document predictors of longitudinal persistence and desistance of TNB gender identity within the subsample of 670 TNB adolescents, and to test whether a three-cluster solution of latent gender incongruence classes resembled a hypothesized typology distinguishing between classical gender dysphoria, autogynephilia, and ROGD. Finite mixture regression identified three latent profiles of gender incongruence, but these appeared to describe mild, moderate, and pronounced subtypes of ROGD rather than the triune typology. Participants in the “mild” cluster were exclusively nonbinary (as opposed to transgender-identified), had a 100% longitudinal desistance rate, sex-typical gender expression, and low levels of internalizing symptoms and social stress. Conversely, participants in the “pronounced” cluster had the highest rates of transgender identification, nonheterosexual orientation, internalizing symptoms, and social stress, and the highest longitudinal persistence rates at 45%. Biopsychosocial correlates of gender incongruence largely did not predict longitudinal TNB persistence or desistance; the strongest predictors were gender identity and sexual orientation.
I conclude by proposing a new biopsychosocial model of adolescent gender dysphoria, synthesizing these findings on adolescent gender incongruence with existing clinical and neurobiological theories of gender dysphoria. This model accounts for recent rises in nonbinary identification among adolescents without gender dysphoria symptoms by introducing a subtype of gender incongruence not captured in existing typologies, characterized by identity exploration, low persistence rates, and sex-typical gender expression and psychosocial adjustment. Sociocultural and neuroendocrine pathways to gender incongruence are discussed.Psycholog
Sensory Feature Representation in Autism: Insights from Naturalistic Neuroimaging
Get PDFSensory-perceptual differences are widely reported in autism, affecting up to 90% of the population, yet the precise underlying neural mechanisms of these observed differences remain poorly characterized and understood. Neuroimaging studies hint at altered sensory cortical activation, but have had mixed findings across different brain regions and often lack theoretical grounding and ecological validity. Naturalistic stimuli, such as movies, have emerged as powerful tools that engage broad neural circuits and permit a more comprehensive mapping of sensory representations in the brain. Applying encoding models to large-scale fMRI datasets of this more ecologically valid data offers a rigorous approach to investigate perceptual differences across both individuals and broader diagnostic groups. Sensory-based theories of autism, such as enhanced perceptual functioning (EPF) and weak central coherence (WCC), posit differences in perceptual styles, such as enhanced low-level sensory processing or a local over global perceptual preference. More broadly across development, perceptual preferences have been observed as shifts in the dominance of sensory modalities. This dissertation employs naturalistic fMRI paired with encoding models to characterize brain-based sensory representations, quantifying the balance of different feature classes represented in the brain and providing a novel and nuanced framework for understanding the neural underpinnings of sensory differences in autism. The first study, Chapter 2, tested the theories of EPF and WCC by contrasting high- vs low-level feature representations during naturalistic movie viewing. This study found that encoding model-derived metrics aligned more with WCC than EPF, with results driven by differences in feature representation in key social and integration brain regions. The second study, Chapter 3, examined sensory dominance by comparing audio vs visual feature representations, and found few differences across autistic and neurotypical groups but widespread age-related effects. Together, these studies demonstrate differences in the balance of high- vs low-level sensory information represented in the brain in autism, but a generally conserved balance of features across audio and visual modalities. These results offer more context into sensory theories of autism and demonstrate the utility of naturalistic movies and encoding models in the context of neurodevelopmental conditions.Speech and Hearing Bioscience and Technolog
From Cancer Initiation to Clinical Insight: Computational and Machine Learning Approaches to Tumor Dynamics and Precision Oncology
No full textDespite extensive research, many fundamental questions in cancer biology and clinical oncology remain unanswered, from the mechanisms underlying cancer initiation and early development to factors that influence diagnosis, treatment response, and patient outcomes. Computational modeling and machine learning frameworks provide powerful tools to address these challenges by enabling large scale mechanistic modeling and the integration of complex molecular and clinical information to uncover mechanisms, generate hypotheses, and guide clinical decisions. This thesis combines computational modeling of cancer initiation and machine learning frameworks for precision oncology to bridge fundamental and translational aspects of cancer research.
Studies of cancer initiation are impeded by complications of identifying and tracking the cell of origin. Recent work has shown that mutagen-induced DNA lesions can persist over multiple rounds of cell division, leaving a statistically interpretable footprint of cancer initiating events. Specifically, it allows estimation of the number of divisions between the DNA lesion introduction and the most recent common ancestor of the developed tumor (LAD). We developed a branching process model of cancer evolution following lesion introduction, and analyzed footprints of segregating lesions from previously published experimental mouse data and post-chemotherapy human metastatic tumors to obtain LAD estimates. We show that in all contexts cancer clones tended to start early, usually within 4 cell generations. Analytical and computational implementations of the branching process model suggested the fitness advantage of early cancer drivers must have exceeded 30% to achieve such early clone initiation.
At the clinical end of the spectrum, precision oncology has informed cancer care by enabling the discovery and application of diagnostic, prognostic, and/or predictive molecular biomarkers. However, many patients lack actionable biomarkers or fail to respond to biomarker-directed therapies. Patient similarity approaches can leverage comprehensive tumor profiling and prior clinical experiences from large cohorts for decision support, facilitating broader realization of precision oncology benefits. We developed a deep learning based modeling framework using real-world clinicogenomic data from a tertiary cancer center to (i) measure patient similarity based on embedded tumor genomic profiles and (ii) evaluate the association of derived patient subgroups and neighborhoods with shared therapeutic outcomes in a breast cancer specific and in a histology-agnostic pan-cancer setting. The model recovered clinically meaningful patient groups of both expected and previously unknown therapeutic associations, as well as patient-specific neighborhoods that could inform therapeutic trajectories more often than expected by chance in multiple clinical contexts. Moreover, model utility extended to patients without actionable genomic biomarkers and those with cancer of unknown primary (CUP) diagnoses, where neighborhoods aligned with independently predicted primary cancer type. These neighborhoods could also be examined over time in a continuously learning scenario.
Overall, these studies integrate fundamental models of cancer evolution and translational machine learning to develop approaches that advance cancer research from onset to clinical decision making. Our branching process model allowed inference of tumor initiation and growth parameters based on events preceding the most recent common ancestor of the initiating clone as opposed to characteristics of fully grown tumors. In parallel, our similarity-based modeling framework distilled complex molecular and clinical data into concise, context-specific insights that augment rather than replace clinician judgment, providing a foundation for real-time learning, patient-centered decision support in precision oncology.Biomedical Informatic