UC San Diego

Motivation: Large pharmacogenomics screens have generated a wealth of information cataloguing the responses of more than a thousand tumor cell-line models to FDA-approved and exploratory drugs. Although centralized repositories have consolidated data access, the diversity of experimental platforms and response metrics used in these screens have made it challenging to integrate and compare their measured drug responses. Towards better pharmacogenomic data harmonization, we surveyed a range of data analysis protocols based on different curve-fitting functions (sigmoid, piecewise linear), different response metrics (IC50, EC50, integrated AUC), and different drug concentration windows (full range or truncated).Results: We found that an AUC derived from a sigmoidal curve fitted to a truncated dose range yields the strongest agreement between screening platforms, significantly bettering other protocols surveyed. This harmonization procedure also best aligns drug responses across successive iterations of the same platform. These findings broadly inform efforts to integrate drug response data in large-scale analyses.Availability and implementation: The source code to generate drug response profiles and correlations are available at https://github.com/digitaltumors/Pharmacogenomics_Screens_Harmonization.git.Supplementary information: Input data for calculating drug response profiles can be downloaded at and . Recalculated drug response profiles from the truncated and full dose range can be downloaded at .

Hypoplastic left heart syndrome is the leading cause of death among infants with congenital heart defects. Through analysis of wild type cells and cells exposed to a mechanically altered in-vivo environment across key developmental stages, we analyze how hemodynamic perturbations alter mechanosensitive cellular and nuclear remodeling.In this study, we assess morphological changes in cytoplasmic and nuclear areas, lamin A/C fluorescence, and yes-associated protein nuclear localization across substrate stiffness, developmental stage, and physiological state. We expose cells to an altered in-vivo mechanical environment, in the form of left-atrial ligation (LAL), and examine subsequent changes in cellular behavior on different substrate stiffnesses. We found that these three factors differentially regulated cell morphology. Mechanical perturbation of the in-vivo environment disrupted mechanosensation across gel stiffness conditions, often reducing sensitivity to developmental stage. The stiff 40 kPa condition caused cells to adopt wider spread triangular morphologies with less sensitivity to developmental stage or LAL due to high extracellular stress. The soft 0.3 kPa condition deviated from the native 1-2 kPa condition due to low extracellular tensile forces. These cells spread uniaxially as they developed, initially adopting more compact conformations than the 1-2 kPa cells but spreading farther at HH31. Lamin A/C intensity trends reflected this behavior, demonstrating relative insensitivity at 40 kPa and a reversal of stage-specific response at 0.3 kPa. An enhanced understanding of mechanosensitive responses across stiffness, stage, and physiological condition allows for better characterization of cardiac disease progression, which can improve detection and serve as a basis for regenerative strategies.

Phthalates (Phthalic Acid Esters; PAEs) are widely used toxic plasticizers, and their contamination in oceans and air is widespread due to their leachability. Previous studies have detected PAEs in air and water in remote marine environments worldwide, suggesting their transfer from the ocean to the atmosphere. However, no prior work has quantified PAE enrichment in the ocean surface layer or in marine aerosols. Here, we investigate the transfer characteristics and enrichment factors (EFs) of three common PAEs: dimethyl phthalate (DMP), dibutyl phthalate (DBP), and di(2-ethylhexyl) phthalate (DEHP), from artificial seawater to an artificial (i.e., palmitic acid) sea surface microlayer (SML) and in nascent sea spray aerosols (SSA). The relative extent of enrichment was found to depend on PAE hydrophobicity and volatility. DBP exhibited the highest EFs in both the artificial SML and nascent SSA compared to other phthalates. DBP EFs in the artificial SML from two independently prepared tanks were 34.3(±1.6) and 20(±3) when present alone and 21.1(±0.9) and 16(±3) when mixed with DMP and DEHP. DMP EFs in the artificial SML were 1.25(±0.10) and 5.45(±0.19) when alone and 1.47(±0.10) and 1.47(±0.07) when mixed. DBP was also significantly more enriched than DMP in nascent SSA, with EFs of 28800(±1300) and 16000(±7000) when alone, and 57400(±2500) and 35000(±8000) when mixed. DMP EFs in SSA were much lower, with values of 6(±12) and 450(±50) when alone and 3(±1) and 542(±24) when mixed. EFs for DEHP were inconclusive due to potential matrix effects, contamination, or degradation. Greater EFs were measured when particles with Dp > 300 nm were absent, suggesting a size dependence in phthalate enrichment in SSA, consistent with other contaminant enrichment studies. These results demonstrate, for the first time, that the sea surface layer and sea spray aerosols may strongly enrich certain PAEs, particularly DBP. This highlights the importance of the ocean surface in mediating the transfer of toxic phthalates from the ocean to the atmosphere.

Electroencephalography (EEG) reports contain clinically important information regarding epilepsy and brain function. However, these findings are typically documented as unstructured free text, limiting their accessibility for large-scale clinical research, quality improvement initiatives, and decision-support systems. We propose a hybrid natural language processing (NLP) pipeline that leverages both fuzzy-matching and open-weight generative large language models (LLMs) to extract relevant features from UCSD Neurology’s highly heterogeneous and time-ambiguous EEG reports. Our best-performing quantized method, Qwen3 14B Q6_K with one-shot quote-extraction prompting, achieved a macro F1 of 0.72 (0.845, micro F1), macro sensitivity of 0.832, specificity of 0.962, and a mean processing report time of 7.2s.

CD8+ T cells are the primary cytotoxic immune cells that are capable of directly killing cancer cells; however, their tumor-killing potential in solid tumors is often limited by insufficient infiltration into the tumor microenvironment. Recent findings suggest that the limited cytotoxic capacity of T cells is partly due to the increased stiffness of the extracellular matrix (ECM) that acts as a physical barrier. This barrier precludes direct T cell-cancer contact, a pre-requisite for contact-dependent killing. Despite this, the mechanisms by which ECM stiffness regulates T cell functions are not fully understood. Furthermore, recent findings demonstrate that tumors are not only stiff, but viscoelastic. Viscoelasticity is a mechanical property reflecting the time-dependent dissipation or relaxation of stresses within a material following deformation. However, the role of viscoelasticity in regulating CD8+ T cell migration remains poorly understood. To address this question, we developed collagen-alginate matrices with matched stiffness but distinct viscoelasticity to investigate how matrix viscoelasticity influences CD8+ T cell migration. Increasing viscoelasticity increased the mean cell volume of CD8+ T cells, decreased the migration straightness, and increased the proportion of the migration phenotype with moderate speed but higher speed standard deviation. The mechanosensitive ion channel inhibitor decreased cell migration speed and straightness in the low-viscoelasticity matrix but caused an opposite effect in the high-viscoelasticity matrix. Additionally, adding the ion channel inhibitor led to a higher proportion of inefficient migration subtype in the low-viscoelasticity matrix but not in the high-viscoelasticity matrix. These suggested that in tissues with lower viscoelasticity, normal mechanosensitive ion channel activity is important for CD8+ T cells to migrate efficiently. Together, these findings highlight the importance of matrix viscoelasticity and mechanosensitive ion channels in regulating CD8+ T cell migration and suggest that understanding and modulating migration phenotypes may provide new strategies to enhance T cell infiltration into solid tumors by engineering Piezo1 activity according to different tissue viscoelasticity.

The decarbonization of the global energy system depends on electrochemical energy storage that is at once higher in energy, faster to charge, safer, and able to operate across a wide temperature range. While conventional lithium-ion batteries (LIBs) have been widely adopted for consumer electronics and, more recently, electric vehicles, graphite-based LIBs remain bounded on two fronts: power and energy density. The first bound is on power density. Through a study of commercial lithium iron phosphate/graphite cells cycled under fast-charging conditions, this work shows that cell failure is driven jointly by the sluggish kinetics of lithium-ion intercalation into graphite and by the continuous growth of the solid electrolyte interphase (SEI), with the balance between these mechanisms shifting with charge rate and temperature. The second bound is on energy: because the usable energy density of the cell is capped by the intercalation chemistry of the graphite anode. As such, lithium metal anode is among the most promising candidates for high- energy-density batteries (> 500 Wh kg-1, > 1000 Wh L-1), owing to its high theoretical capacity, low reduction potential, and low density, but its commercialization is limited by short cycle life from continuous dendrite growth and by safety concerns arising from porous electrodeposition. As a common denominator in battery systems, electrolyte engineering remains the most direct means of addressing these issues, simultaneously governing the rate of lithium-ion transport through the bulk and across the interface and dictating the chemistry of the anode interphase. Yet the two objectives are often in tension: high-concentration and localized-high concentration electrolytes form favorable, anion-derived interphases but do so at the cost of high viscosity, reduced ionic conductivity, and poor wettability of thick electrodes, underscoring the need for an electrolyte that delivers fast transport and a stable interphase concomitantly.These limits motivate the central strategy of this dissertation: the design of liquefied gas electrolytes (LGEs), whose low viscosity, wide liquid range, and weakly coordinating character relax transport limitations while stabilizing reactive metal surfaces. Building on this approach, an aggregate-rich electrolyte that combines the high-voltage stability of an ionic liquid with the fluidity of the liquefied gaseous solvent, fluoromethane (FM), is shown to enable a thin lithium metal anode to cycle stably against a high-voltage nickel-rich cathode, retaining high conductivity and low-temperature operation.The same design principle is then generalized from lithium to sodium. Sodium is attractive for its natural abundance, and its lower charge density yields comparatively weak ion–solvent binding that pairs naturally with the weakly coordinating nature of liquefied gas solvents; exploiting this complementarity, a LGE is developed based on a similar concept of localized highly concentrated electrolytes that lowers the sodium-ion desolvation barrier and sustains fast transport, enabling stable sodium metal cycling deep into low-temperature regimes. Taken together, these studies establish that rationally engineered LGEs can unlock high-energy density systems with the use of alkali metal anodes while maintaining competitive power density and rate capability.

Plastics are highly inert and environmentally persistent petroleum derivatives employed in commercial, industrial, and residential applications. Disintegration via mechanical and chemical means generates plastic fragments of micro (≤ 1 mm) or nanoscopic (≤ 1 μm) sizes. Such particles are highly heterogeneous in size, morphology, structural chemistry, and surface topography. With a vast and often indiscernible environmental burden, the biochemical and biophysical influences of plastics on mammalian physiology are of concern, especially regarding chronic inflammation. Utilizing three monocyte-derived model systems, a mouse cell line, bone marrow-derived macrophages (BMDMs), and patient-derived, monocyte-enriched PBMCs, morphologically heterogeneous microplastics, such as polystyrene (PS) and polyacrylonitrile (PAN), at low, medium, and high dosages were imaged for topographical variety via scanning electron microscopy (SEM), size characterized by dynamic light scattering (DLS), analyzed for chemical composition by MALDI-TOF and FTIR, and screened in a TNF-ɑ ELISA. The primary thrusts are to observe hallmark inflammatory responses to microplastic exposure, profile the physicochemical structure of polymers, and reproducibly and sterilely generate and characterize microplastics from larger fragments through sonication, microwaving, and mechanical degradation. All three cell types responded positively to PAN at sizes <1 μm and at high concentrations, with the dual upregulation of TNF-ɑ and M2-like transcriptional programs in murine cells. Two PAN vendors differed in diameter, average molecular weight, sonication receptivity, fibrousness, porosity, and FTIR-derived chemical composition, with time-dependent aggregation tendencies. These data suggest that a microplastic derived from PAN can have biological effects on macrophages, promoting a tissue remodeling signature with preserved, antithetical mechanisms along the TNF-ɑ axis. Future directions will include performing a mechanistic evaluation, incubating size-enriched PAN subpopulations, and screening for unknown soluble contaminants with inflammatory potential.

Transcription factors (TFs) are primary regulators of cell identity, and understanding how their disruption affects cell state is a core objective of functional genomics. Perturb-seq, which combines gene perturbations with single-cell transcriptomic readouts, enables high-throughput measurement of the downstream responses to perturbations. This work reanalyzes two published genome-scale Perturb-seq datasets - one in KOLF2.1J human induced pluripotent stem cells, one in the K562 leukemia cell line - using a single analytical workflow. Building upon these studies, this work interprets the downstream programs each TF knockdown engages, at the level of the whole dataset, individual clusters, and single knockdowns. In both datasets, TF knockdowns reshaped cell states along lineage-associated axes. In KOLF2.1J, disruption of the pluripotency network destabilized the pluripotent cell state and revealed a bias towards a neuroectoderm lineage. In K562, TF knockdown responses highlighted distinct regulators of the erythroid and myeloid programs. Lineage specification emerged as the clearest biological signal above a generic TF knockdown program in both datasets. Cross-dataset comparison further identified regulators whose transcriptional responses depended on the cell state, providing a demonstration of context-dependent regulation across two cell lines with differing lineage potential. Together, these results characterize how TF knockdowns engage and destabilize lineage specification programs. Further, this work demonstrates that published Perturb-seq datasets yield interpretable biology beyond the original analysis when the downstream effects of perturbations are examined directly.

Metastatic pancreatic ductal adenocarcinoma (PDAC) is caused by invasive cell clusters that collectively migrate from the primary tumor to distant tissue. Efforts to characterize this invasion phenotype rely on assays to measure functional morphology or prognostic tumor markers, both of which have been shown to be regulated by the matrix microenvironment via physical properties such as stiffness, confinement, and degradability. However, it is unclear how physical properties intersect with the unique metabolic reprogramming of cancer cells within these environments to contribute to a widely heterogeneous array of phenotypes within the same tumor. In particular, high density collagen I (HDC) induces heterogenous collective cell morphology in vitro when seeded as single cells, which grow into multicellular structures with a range of collective morphologies depending on invasive phenotype. By systematically varying the density of fibrillar collagen I, we first show that collective invasion is modulated by matrix architecture, and that this invasive phenotype is more prominently modeled in HDC than in gold-standard basement membrane extract (Matrigel), but not through suppression of epithelial traits. Using 3D HDC gels as invasion assays, we then pharmacologically inhibited oxidative phosphorylation, which reduced the frequency of collectively invasive structures but not their proliferation. Interestingly, categorizing multicellular structures by morphology reveals that proliferation is correlated with invasive phenotype. Finally, we show that PDAC that is resistant to KRAS inhibitors is also more collectively invasive in collagen I.

Enabling the next generation of rechargeable batteries necessitates the fabrication and manufacturing of energy dense electrode materials at scale. Sodium metal is an attractive option because of its abundance and potential cost advantages relative to lithium. As such, sodium-metal has been widely researched as a next-generation anode for rechargeable batteries. However a central issue in sodium-metal research is the lack of a universal sodium source for laboratory experiments. Moreover, its mechanical softness leads to processing difficulty which makes for inconsistent studies. In practice, approaches to prepare anodes at lab scale include rolling sodium onto a current collector from sodium ingots, vendor-supplied sodium chips, and in situ electroplating sodium, each of which introduces different structural and practical uncertainties.This thesis examines sodium-metal sources beyond bulk morphology to determine whether they exhibit crystallographic differences that influence electrochemical behavior. A reproducible cell assembly workflow was first established to reduce variability from externally applied stack pressure. Cryogenic focused ion beam/scanning electron microscopy (cryo-FIB/SEM) showed that ingot-derived, chip-based, and electroplated sodium all appeared dense and morphologically similar at the mesoscale. However, wide-angle X-ray scattering (WAXS) revealed clear source dependent differences in crystallographic texture. Sodium chips were closest to a powder-like pattern, while electroplated and mechanically processed sodium showed increasing preferred orientation. Varying electroplating current density further demonstrated current-density-dependent crystallographic texture. Rate testing showed improved high-rate utilization for electroplated sodium relative to sodium chips, supporting crystallography as a critical variable in sodium-metal electrochemistry.