<p dir="ltr">Despite the remarkable advancements of machine learning (ML) technologies in biomedical research, especially in tackling complex human diseases such as cancer and Alzheimer's disease, a considerable gap persists between promising theoretical results and dependable clinical applications in diagnosis, prognosis, and therapeutic decision-making. One of the primary challenges stems from the absence of large high-quality patient datasets, which arises from the cost and human labor required for collecting such datasets and the scarcity of patient samples. Moreover, the inherent complexity of the data often leads to a feature space dimension that is large compared with the sample size, potentially causing instability during training and unreliability in inference. To address these challenges, the transfer learning (TL) approach has been embraced in biomedical ML applications to facilitate knowledge transfer across diverse and related biological contexts. Leveraging this principle, we introduce an unsupervised multi-view TL algorithm, named MVTOT [1], which enables the analysis of various biomarkers across different cancer types. Specifically, we compress high-dimensional biomarkers from different cancer types into a low-dimensional feature space via nonnegative matrix factorization and distill common information shared by various cancer types using the Wasserstein distance defined by Optimal Transport theory. We evaluate the stratification performance on three early-stage cancers from the Cancer Genome Atlas (TCGA) project. Our framework, compared with other benchmark methods, demonstrates superior accuracy in patient survival outcome stratification.</p><p dir="ltr">Additionally, while patient-level stratification has enhanced clinical decision-making, our understanding of diseases at the single-cell (SC) level remains limited, which is crucial for deciphering disease progression mechanisms, monitoring drug responses, and prioritizing drug targets. It is essential to associate each SC with patient-level clinical traits such as survival hazard, drug response, and disease subtypes. However, SC samples often lack direct labeling with these traits, and the significant statistical gap between patient and SC-level gene expressions impedes the transfer of well-annotated patient-level disease attributes to SCs. Domain adaptation (DA), a TL subfield, addresses this challenge by training a domain-invariant feature extractor for both patient and SC gene expression matrices, facilitating the successful application of ML models trained on patient-level data to SC samples. Expanding upon an established deep-learning-based DA model, DEGAS [2], we substitute their computationally ineffective maximum mean discrepancy loss with the Wasserstein distance as the metric for domain discrepancy. This substitution facilitates the embedding of both SC and patient inputs into a common latent feature space. Subsequently, employing the model trained on patient-level disease attributes, we predict SC-level survival hazard, disease status, and drug response for prostate cancer, Alzheimer's SC data, and multiple myeloma data, respectively. Our approach outperforms benchmark studies, uncovering clinically significant cell subgroups and revealing the correlation between survival hazard and drug response at the SC level.</p><p dir="ltr">Furthermore, in addition to these approaches, we acknowledge the effectiveness of TL and image analysis in stratifying patients with early and late-stage Mild Cognitive Impairment based on neuroimaging, as well as predicting survival and metastasis in melanoma based on histological images. These applications underscore the potential of employing ML methods, especially TL algorithms, in addressing biomedical issues from various angles, thereby enhancing our understanding of disease mechanisms and developing new biomarkers predicting patient outcomes.</p>

Initial attachment to the extracellular matrix (ECM) and consequent spreading is a necessary process in the cell cycle of which little is known. Cell spreading has been well-recognized in 2D systems, however, the native fibrous ECM presents cells with 3D biophysical cues. Thus, using suspended fibers as model systems, we present the development of a novel platform (Cell-STEPs) capable of capturing cell attachment dynamics and forces from the moment a cell in suspension contacts the fiber. Cell-STEPs comprises of a custom glass-bottom petri dish with a lid to deliver a constant supply of CO2 to maintain pH. Fibrous scaffolds are attached in the dish to allow cellular investigations over extended periods of time. We find that cell-fiber attachment occurs in three progressive phases: initial attachment of cell to fiber (phase 0), rapid drop in circularity (phase 1), and increase in cell spread area (phase 2). Furthermore, using iterative inverse methods, forces involved in cell spreading through deflection of fibers were estimated. Our findings provide new insights in attachment biomechanics, including initial sensing and latching of cell to fiber with a negligible or protrusive force, followed by rapid loss in circularity through protrusion sensing at nearly constant spread area and minimal force generation, transitioning to a final phase of increased contractile forces until spread area and force saturation is observed. Also, anisotropic spreading of cells on single and two-fibers are closely related, while cells attached to several fibers take longer and spread isotropically. The Cell-STEPs platform allows, for the first time, detailed interrogations in the discrete and orchestrated adhesion steps involved in cell-fibrous matrix recognition and attachment along with simultaneous measurements of forces involved in cell attachment. / Master of Science

Extracellular matrix (ECM) is a fibrous natural cell environment, possessing complicated micro-and nano- architectures, which provides signaling cues and influences cell behavior. Mimicking this three dimensional environment in vitro is a challenge in developmental and disease biology. Here, suspended multilayer hierarchical nanofiber assemblies fabricated using the non-electrospinning STEP (Spinneret based Tunable Engineered Parameter) fiber manufacturing technique with controlled fiber diameter (microns to less than 100 nm), orientation and spacing in single and multiple layers are demonstrated as biological scaffolds. Hierarchical nanofiber assemblies were developed to control single cell shape (shape index from 0.15 to 0.57), nuclei shape (shape index 0.75 to 0.99) and focal adhesion cluster length (8-15 micrometer). To further investigate single cell-ECM biophysical interactions, nanofiber nets fused in crisscross patterns were manufactured to measure the "inside out" contractile forces of single mesenchymal stem cells (MSCs). The contractile forces (18-320 nano Newton) were found to scale with fiber structural stiffness (2 -100 nano Newton/micrometer). Cells were observed to shed debris on fibers, which were found to exert forces (15-20 nano Newton). Upon CO? deprivation, cells were observed to monotonically reduce cell spread area and contractile forces. During the apoptotic process, cells exerted both expansive and contractile forces. The platform developed in this study allows a wide parametric investigation of biophysical cues which influence cell behaviors with implications in tissue engineering, developmental biology, and disease biology. / Master of Science

Cells in the body interact with a predominantly fibrous microenvironment and constantly adapt to changes in their neighboring physiochemical environment, which has implications in developmental and disease biology. A myriad of in vitro platforms including 2D flat and 3D gel substrates with and without anisotropy have demonstrated cellular alterations to subtle changes in topography. Recently, our work using suspended fibers as a new in vitro biological assay has revealed that cells are able to sense and respond to changes in fiber curvature and structural stiffness as evidenced by alterations to cytoskeleton arrangement, including focal adhesion cluster lengths and nucleus shape indices, leading to altered migration speeds. It is hypothesized that these behaviors occur due to modulation of cellular inside-out forces in response to changes in the external fibrous environment (outside-in). Thus, in this study, we investigate the role of fiber curvature and structural stiffness in force modulation of single cells attached to suspended fibers. Using our previously reported non-electrospinning Spinneret based Tunable Engineered Parameters (STEP) fiber manufacturing platform, we present our findings on single cell inside-out and outside-in forces using fibers of three diameters (250 nm, 400 nm and 800 nm) representing a wide range of structural stiffness (3-45 nN/μm). To investigate cellular adaptability to external perturbation, we present the development of a first-of-its-kind force measurement 'nanonet' platform capable of investigating cell adhesion forces in response to symmetric and non-symmetric (injury model) loading. Our combined findings are multi-fold: (i) Cells on suspended fibers are able to form focal adhesion clusters approximately four times longer than those on flat substrates, which gives them potential to double their migration speeds, (ii) Nanonets as force probes show that the contractility-based inside-out forces are nearly equally distributed on both sides of the cell body, and that overall force magnitudes are dependent on fiber structural stiffness, and (iii) External perturbation can evenly (symmetric) or unevenly (non-symmetric) distribute forces within the cell, and the resulting bias causes diameter-dependent outside-in adhesion force response. Finally, we demonstrate the power of the developed force measurement platform by extending our studies to cell-cell junctional forces as well as single-cell disease models including cancer and aortic aneurysm. / Ph. D.

Recent severe weather research has examined the potential role of total lightning patterns in the severe thunderstorm warning-decision process although none to-date have examined these patterns in explicitly weak-shear environments. Total lightning flashes detected by the Earth Networks Total Lightning Network (ENTLN) during the 2012-13 convective seasons (1 May – 31 August) over a region of the Central Appalachian Mountains were clustered into likely discrete thunderstorms and subsequently classified as either single-cell or multicell/supercell storm modes. The classification of storms was determined using a storm index (SI) which was informed by current National Weather Service (NWS) identification techniques. The 36 days meeting the minimum threshold of lightning activity were divided into 24 lightning-defined (LD) single-cell thunderstorm days and 12 LD multicell/supercell days. LD single-cell days possessed statistically significant lower 0000 UTC 0-6 km wind shear (13.8 knots) than LD multicell/supercell days (26.5 knots) consistent with traditional expectations of single-cell and multicell/supercell environments respectively. The popular 2σ total lightning jump algorithm was applied to all flashes associated with 470 individual LD thunderstorms. The frequencies of the storms’ total lightning jumps were then compared against any associated severe weather reports as an accuracy assessment. The overall performance of the algorithm among both categories was much poorer than in previous studies. While probability of detections (POD) of the 2σ algorithm were comparable to previous research, false alarm rates (FAR) were much greater than previously documented. Given these results, the 2σ algorithm does not appear fit for operational use in a weak shear environment. / Master of Science

From the discovery of developmental gradients to pioneering some of the first gene regulatory models, the sea urchin model has played a foundational role in deciphering the complex molecular mechanisms behind the phenomena that underlie pattern formation during embryonic development. Of particular interest to our lab, primary mesenchyme cells (PMCs), a skeletogenic lineage, provide an excellent system for understanding the mechanisms behind skeletal pattern formation. Sea urchin skeletal patterning is driven by ectodermal cues that are differentially expressed in space and time; these cues instruct the PMCs. Originating as a homogeneous population, PMCs diversify in response to patterning cue reception, then produce distinct skeletal elements as a function of the cues that they have received from the ectoderm. However, the exact mechanisms underpinning PMC diversification and the role that individual ectodermal cues play to mediate this diversification process is poorly understood. To bridge that knowledge gap, this work leverages multiple data modalities, including single-cell RNA sequencing (scRNA-seq) and 3D visualization of gene expression in normal and perturbed embryos to not only present an exhaustive description of PMC diversification, but also offers novel computational approaches and the development of resources necessary for these studies. First, we present the novel algorithm ICAT. Created to correctly identify cell states from mixedcondition scRNA-seq experiments, ICAT plays a necessary role in identifying PMC subpopulations affected by ectodermal cue disruption. Using simulated and real datasets, we benchmark ICAT against several state-of-the-art workflows, and find ICAT provides more robust and sensitive performance compared to current practices. We further validate ICAT in vivo using single molecule fluorescent in situ hybridization (FISH) and show that, compared to leading algorithms, ICAT uniquely and correctly characterizes the effects of patterning cue disruption on PMC subpopulation composition. Finally, by combining temporal scRNA-seq data throughout skeletal patterning with a newly generated spatial gene expression reference map, we not only identify distinct PMC subpopulations, but also provide spatial and temporal coherence to each of their developmental trajectories during skeletal pattern formation. We compliment this work by inferring the gene regulatory networks underlying PMC diversification and thereby identifying the transcriptional regulators that function as network hubs. We empirically demonstrate that these hubs are required for skeletal patterning, and spatially map their expression within the PMCs. Sequencing single PMCs isolated from embryos in which ectodermal cue function was inhibited, we show that functional loss of each cue uniquely disrupts the PMC gene regulatory network and characterize the subsequent compositional effects of PMC subpopulations. Taken together, this work defines the spatiotemporal details of PMC diversification in normal embryos as well as in embryos with individual cue losses, as well as offering numerous novel computational methods and resources necessary for these advances. / 2026-09-26T00:00:00Z

Since the emergence of biology as a quantitative science in the past century, a lot of biological discoveries have been driven by milestone technical advances such as X-ray crystallography, fluorescence microscopy and high-throughput sequencing. Fluorescence microscopy is widely used to explore the nanoscale cellular world because of its superb sensitivity and spatial resolution. However, many species (e.g. lipids, small proteins) are non-fluorescent and are difficult to label without disturbing their native functions. In the first part of the dissertation, we explore using three different contrast mechanisms for label-free imaging of these species – absorption and stimulated emission (Chapter 2), heat generation and diffusion (Chapter 3) and nonlinear scattering (Chapter 4). We demonstrate label-free imaging of blood vessels, cytochromes, drugs for photodynamic therapy, and muscle and brain tissues with three dimensional optical sectioning capability. With the rapid development of high throughput genotyping techniques, genome analysis is currently routinely done genome-wide with single nucleotide resolution. However, a large amount of starting materials are often required for whole genome analysis. The dynamic changes in DNA molecules generate intra-sample heterogeneity. Even with the same genome content, different cells often have very different transcriptome profiles in a functional organism. Such intra-sample heterogeneities in the genome and transcriptome are often masked by ensemble analysis. In this second part of the dissertation, we first introduce a whole genome amplification method with high coverage in sequencing single human cells (Chapter 6). We then use the technique to study meiotic recombinations in sperm cells from an individual (Chapter 7). We further develop a technique that enables digital counting of genome fragments and whole genome haplotyping in single cells (Chapter 8). And we introduce our ongoing efforts on single cell transcriptome analysis (Chapter 9). In the end, we introduce our initial effort in exploring the genome accessibility at the single cell level (Chapter 9). Through the development of techniques probing the single cell genome, transcriptome and possibly epigenome, we hope to provide a toolbox for studying biological processes with genome-wide and single cell resolution. / Chemistry and Chemical Biology

Anticorpos monoclonais (AcMos) para uso terapêutico correspondem a uma área importante na indústria de biofármacos, em especial os AcMos humanos, que apresentam menor probabilidade de elicitar imunogenicidade. O objetivo deste trabalho consistiu em obter AcMos humanos antitetânicos através da separação de linfócitos B produtores de anticorpos específicos utilizando o antígeno ou de plasmablastos. As células foram coletadas de doadores após vacinação e separadas por equipamento de cell sorter. As regiões variáveis dos anticorpos foram amplificadas e clonadas em vetores de expressão, que foram usados para transfectar transitoriamente células HEK293-F. O uso da toxina tetânica conjugada independentemente com dois marcadores, biotina e Alexa Fluor® 647, possibilitou a separação específica de linfócitos B produtores de AcMos antitetânicos, que foram avaliados por ELISA, western blotting e pela inibição da ligação da toxina ao gangliosídio GT1b. O ensaio in vivo mostrou proteção total dos animais contra a toxina tetânica quando três AcMos foram usados em conjunto. / Monoclonal antibodies (mAbs) for therapeutic use correspond to a major area of the biopharmaceutical industry, especially human mAbs that are less prone to elicit immunogenicity. The objective of this work was to obtain anti-tetanus human mAbs through separation of memory B lymphocytes producing specific antibodies stained with the antigen or plasmablasts. Cells were collected from peripheral blood of donors after vaccination and separated through cell sorting. The variable regions of the antibodies were amplified and cloned in expression vectors for transient transfection of HEK293-F cells. The staining with the tetanus toxin labeled independently with two markers, biotin and Alexa Fluor® 647 allowed the separation of specific B lymphocytes producing anti-tetanus mAbs. The antibodies expressed were evaluated by ELISA, western blotting and the inhibition of the binding of the tetanus toxin to the ganglioside GT1b. The in vivo neutralization assay showed that a pool of three different mAbs were able to protect mice against the tetanus toxin.

Anticorpos monoclonais (AcMos) para uso terapêutico correspondem a uma área importante na indústria de biofármacos, em especial os AcMos humanos, que apresentam menor probabilidade de elicitar imunogenicidade. O objetivo deste trabalho consistiu em obter AcMos humanos antitetânicos através da separação de linfócitos B produtores de anticorpos específicos utilizando o antígeno ou de plasmablastos. As células foram coletadas de doadores após vacinação e separadas por equipamento de cell sorter. As regiões variáveis dos anticorpos foram amplificadas e clonadas em vetores de expressão, que foram usados para transfectar transitoriamente células HEK293-F. O uso da toxina tetânica conjugada independentemente com dois marcadores, biotina e Alexa Fluor® 647, possibilitou a separação específica de linfócitos B produtores de AcMos antitetânicos, que foram avaliados por ELISA, western blotting e pela inibição da ligação da toxina ao gangliosídio GT1b. O ensaio in vivo mostrou proteção total dos animais contra a toxina tetânica quando três AcMos foram usados em conjunto. / Monoclonal antibodies (mAbs) for therapeutic use correspond to a major area of the biopharmaceutical industry, especially human mAbs that are less prone to elicit immunogenicity. The objective of this work was to obtain anti-tetanus human mAbs through separation of memory B lymphocytes producing specific antibodies stained with the antigen or plasmablasts. Cells were collected from peripheral blood of donors after vaccination and separated through cell sorting. The variable regions of the antibodies were amplified and cloned in expression vectors for transient transfection of HEK293-F cells. The staining with the tetanus toxin labeled independently with two markers, biotin and Alexa Fluor® 647 allowed the separation of specific B lymphocytes producing anti-tetanus mAbs. The antibodies expressed were evaluated by ELISA, western blotting and the inhibition of the binding of the tetanus toxin to the ganglioside GT1b. The in vivo neutralization assay showed that a pool of three different mAbs were able to protect mice against the tetanus toxin.

Das Herz ist das zentrale Kreislauforgan in unserem Körper und jede Abweichung seiner Funktion wirkt sich negativ auf die Homöostase des gesamten Körpers aus. Die Herzfunktion beruht auf der Synergie der Zellen, die das Organ bilden. Die detaillierte zelluläre Zusammensetzung sowie die Funktionalität der einzelnen Zellen müssen noch ermittelt werden, und diese Arbeit ist eine wichtige Ergänzung dieser Bemühungen. Dank der jüngsten Entwicklungen in den Einzelzelltechnologien sind wir nun in der Lage, Transkriptome einzelner Zellen aus komplexem Gewebe in beispiellosem Umfang zu charakterisieren. Im ersten Schritt eines solchen Experiments müssen die Zellen und Zellkerne aus dem Gewebe befreit und vereinzelt werden. Herzgewebe wirft in dieser Hinsicht einzigartige Herausforderungen auf, darunter die Knappheit des gesunden menschlichen Herzgewebes für die Forschung, das Vorhandensein von Kardiomyozyten, die aufgrund ihrer Größe nicht durch Microfluid-basierte Standardinstrumente passen und deren Multinukleation, sowie mögliche Voreingenommenheit verschiedener Methoden zur Gewebedissoziation. Hier präsentiere ich den umfassenden Zellatlas des gesunden erwachsenen menschlichen Herzens. Ich beginne mit der Methodenentwicklung zur Isolierung von einzelnen Zellen und Zellkernen aus Mausherzen. Um den Zellatlas des menschlichen Herzens zu erstellen, analysiere ich einen Datensatz von fast einer halben Million Einzelzellen und Zellkerne aus sechs Herzregionen von vierzehn gesunden Menschen. In diesem Atlas definieren wir 11 Hauptzelltypen und 62 Zellzustände des menschlichen Herzens. Ein tieferer Fokus wird auf das Herzgefäßsystem gelegt und die Zellen der arterio-venösen Achse sowie deren Wechselwirkungen und potenzielle Funktionalität werden definiert. Insgesamt präsentiert diese Dissertation einen komplex Datensatz aus menschlichem Herzgewebe und liefert neue Einblicke in die Biologie des gesunden Herzens mit Implikationen für kardiovaskuläre Erkrankungen. / The heart is the central circulatory organ in our bodies and any discrepancies of its function relative to healthy homeostasis negatively impact the whole body. Cardiac function relies on the synergy of all the cells that constitute the organ. The detailed cellular composition as well as the heterogeneity and functionality of the individual cells is yet to be established and this work is a major advance in this effort. Thanks to the recent developments in single cell genomics technologies, we are now able to profile transcriptomes from individual cells of complex tissues at unprecedented scale. In the first step of such an experiment, the single cells and nuclei need to be liberated from the tissue. Heart tissue presents a unique set of challenges in this regard, including the scarcity of healthy human cardiac tissue for research, large cardiomyocytes that do not fit into the standard droplet-based instruments, multinucleation of cardiomyocytes that might skew the proportions of the recovered nuclei as well as potential bias of tissue dissociation methods. Here I present a cell atlas of the free walls, apex and septum of the healthy adult human heart. I start with methods development for the isolation of single cells and single nuclei from mouse heart. Next, I move to the building of the atlas of the human cells and nuclei, where I describe the dataset of close to half a million single cells and nuclei sampled from 14 organ donors, defining 11 major cell types and 62 cell states of the heart. A deeper focus on the cardiac vasculature defined the cells of the arterio-venous axis as well as their interactions and potential functionality. Overall, this thesis presents a joined dataset of single cells and single nuclei from human cardiac tissues and provides new insights into cardiac biology in heath with implications for cardiovascular disease.