Interdisciplinary Center for Quantitative Modeling in Biology

01

Brady Berg

Abstract:

A repeated pattern in cytoskeletal architecture is the aster, in which a number of F-actin filaments emerge star-shaped from a central node. Aster-based structures occur in cytoplasmic actin, the early stages of the cytokinetic ring in yeast, and in the context of biomimetic materials engineering. In this work, we use computational simulation to show that there is an optimal number of filaments per aster—what we term the "astral number"—that maximizes rigidity, even at a fixed density of F-actin. This nonlinear dependence holds for both the shear and extensional moduli. Furthermore, we find that increasing astral number leads to dramatic increases in the sample-to-sample variability in network rigidity. We explain both effects using percolation theory, wherein the probability that a given network is productively connected exhibits a sharp dependence on parameters. The dependence of network rigidity on astral number may suggest a mechanism by which cells tune the physical properties of their actin networks and may inform efforts to create adaptive synthetic metamaterials inspired by actin networks.

02

Jonathan Castillo

Abstract: Immunotherapy has emerged as a breakthrough in the improvement of survival outcomes for lung adenocarcinoma (LUAD), however effective response requires the combination of blocking inhibitory signals on the tumor surface and antigen presentation to the tumor surface for proper immune recognition. Several commercially available and robust methods exist for identification of tumors displaying immune-inhibitory surface receptors, such as PD-L1, however it is currently difficult to predict effectiveness of antigen presentation on the cell surface. To address this, we utilized clinical and next-generation sequencing data from The Cancer Genome Atlas (TCGA) to identify gene signatures that are correlated to tumor mutational burden (TMB) within cancers of epithelial origins as a surrogate for neoantigen signatures. We identified LINC00261 as a top gene correlated to TMB, whose expression activates DNA damage response pathways in vitro along with resistance to cisplatin. LINC00261 expression was also significantly correlated to MHC class I and II genes involved in endogenous neoantigen presentation expression within the TCGA-LUAD cohort. This relationship was confirmed in vitro through ectopic reintroduction of LINC00261 for key MHC class II presentation genes. Interferon gamma-induced MHC gene activation in vitro was also able to induce endogenous expression of LINC00261. Multiplex immunofluorescence and Xenium In Situ of primary human lung cancer sections suggested that loss of LINC00261 is associated with an immunosuppressive tumor microenvironment. Taken together, our results suggest there is a mechanistic relationship in the silencing of LINC00261 in LUAD and compromised DNA repair, accumulation of mutations, and reduced antitumor immune response.

3

Vincent Cerbantez Bueno

Abstract: The development and growth of multicellular organisms relies on a balance between cell division and differentiation. At the same time, these two processes are regulated by gene activation and repression. In plants, meristems provide the cell source to develop the different organs and tissues and their maintenance is essential for their indeterminate growth. In Arabidopsis, WUSCHEL (WUS) plays a critical role in the shoot apical meristem (SAM) maintenance. This homeodomain transcription factor is synthesized in the rib meristem (RM), then migrates to the central zone and forms a concentration gradient across this diffusion. In low concentrations, WUS acts as a monomer and binds to CLV3 cis-regulatory elements activating its expression. At high concentrations WUS forms homodimers and binds to repress CLV3 expression. At the same time, the synthesized CLV3 peptide restricts the transcription and promotes nuclear accumulation of WUS. Then, somehow WUS self-regulates through CLV3 activation or repression. 

A previously reported cis-regulatory module (CRM) has partially described the mechanism underlying the unique WUS-mediated CLV3 regulation. However, the complete mechanism is still not well understood. Using in silico and molecular analysis, we identified two additional CRM (CRM2 and CRM3), located in the 3’ region of CLV3, where WUS was able to bind with. Our genetic data suggests an essential role of these two new CRM in the regulation of CLV3 levels and spatial expression in the deeper layers of the SAM. Additionally, they seem to cooperate between them and the previously described CRM1.

04

Handan Cetin

Abstract: Stromal and immune cells in the tumor microenvironment undergo extensive metabolic reprogramming that supports cancer progression. We investigated these metabolic adaptations by reconstructing cell type-specific genome-scale metabolic models (GEMs) through integration of single-cell transcriptomics data from Qi et al.'s comprehensive colorectal cancer study, which previously identified correlation patterns between FAP+ fibroblasts and SPP1+/MARCO+ macrophages associated with poor patient outcomes. We reconstructed metabolic models for both normal and tumor-associated phenotypes across multiple subtypes: FGFR2+, FAP+, and CD73+ fibroblasts; DES+ and MFAP5+ myofibroblasts; ICAM1+ and ICAM1- telocytes; and THBS1+, VCAN+, and MARCO+ (SPP1+) macrophages. We extended the generic Human-GEM model with a reactive oxygen species (ROS) module to better capture redox metabolism before generating cell-type specific GEMs, then sampled the solution spaces of individual models. Statistical comparison of metabolic networks and flux solutions revealed significant reprogramming of key pathways in tumor-associated phenotypes. FAP+ fibroblasts showed increased activity in amino acid metabolism, fatty acid oxidation, and carnitine shuttle pathways. Our analysis also revealed that ICAM1+ telocytes exhibited metabolic profiles most similar to FAP+ fibroblasts, suggesting a potential developmental relationship between these subtypes. MARCO+ macrophages exhibited higher activity in arachidonic acid metabolism, pyrimidine metabolism, and alanine, aspartate, and glutamate metabolism. Interestingly, while conventional network metrics failed to distinguish between tumor and normal metabolic networks, geometric analysis revealed different scales at which metabolic reorganization was most pronounced, with each cell type exhibiting a unique value where the tumor-normal distance was maximized. This suggests that metabolic reprogramming occurs at different organizational scales depending on cell type, highlighting the complexity of TME adaptation and the need for higher resolution cell classification. These findings provide insights into the metabolic adaptations that support cellular function within the TME and suggest potential targets for therapeutic intervention that could disrupt tumor-supporting metabolic activities in the colorectal cancer microenvironment.

05

Sakshar Chakravarty

Abstract: Repetitive regions in eukaryotic genomes often contain important functional or regulatory elements. Despite significant algorithmic and technological advancements in genome sequencing and assembly over the past three decades, modern de novo assemblers still struggle to accurately reconstruct highly repetitive regions.

In this work, we introduce RAmbler (Repeat Assembler), a reference-guided assembler specialized for the assembly of complex repetitive regions exclusively from PacBio HiFi reads. RAmbler (i) identifies repetitive regions by detecting unusually high coverage regions after mapping HiFi reads to the draft genome assembly, (ii) finds single-copy k-mers from the HiFi reads, (i.e., k-mers that are expected to occur only once in the genome), (iii) uses the relative location of single-copy k-mers to barcode each HiFi read, (iv) clusters HiFi reads based on their shared barcodes, (v) generates contigs by assembling the reads in each cluster, and (vi) generates a consensus assembly from the overlap graph of the assembled contigs.

Here we show that RAmbler can reconstruct human centromeres and other complex repeats to a quality comparable to the manually-curated telomere-to-telomere human genome assembly. Across over 250 synthetic datasets, RAmbler outperforms hifiasm, LJA, HiCANU, and Verkko across various parameters such as repeat lengths, number of repeats, heterozygosity rates and depth of sequencing.

06

Shiyuan Chen

Abstract: Traditional biology large language models have been based on BERT; they are resource inefficient, undertrained and mostly dedicated to one single task without natural language support. We present RNASeek, a deepseekR1 based multi modal model trained on about 11.7 billions of nucleotides for sequence, structural and natural language modeling. We demonstrate that by utilizing this one single largest model, multiple tasks could be integrated into one by just changing the prompt and with the additional modalities such as structural and genomic feature annotations, our model often requires less training cycle and achieves better accuracy than traditional smaller models on tasks such as stability, efficiency prediction.

07

Lynne Cherchia

Abstract: The prolactin receptor (PRLR) is a single-pass transmembrane receptor driving pancreatic β-cell proliferation via JAK/STAT signaling activation. This signal transduction pathway enables insulin-secreting β-cells to adapt to metabolic stress; however, the precise mechanisms underlying the pathway’s proliferative effect remain ill-defined. Here we implement a pipeline that uses live-cell fluorescence imaging, reconstitution approaches, and fluorescence correlation spectroscopy (FCS) to inform a mathematical model of PRLR signaling in β-cells and build a quantitative, mechanistic understanding of the signaling network. PRLR signaling is dynamic, involving changes in the spatial organization of signaling molecules. We have observed PRLR undergoing rapid internalization, a behavior that has been shown and modeled in other signaling pathways but has not been considered in a mathematical model of PRLR signaling. Such a model is useful for predicting strategies to modulate β-cell function. PRLR internalization is observed in both our minimal engineered PRLR expression system and in native pancreatic tissue, while FCS and chemigenetic labeling with SNAP-tag confirm the presence of a low concentration plasma membrane pool of PRLR. Our imaging data are used to integrate PRLR trafficking dynamics into an ordinary differential equation (ODE) model of PRLR signaling. We employ the ODE model to test hypotheses targeting how the spatial heterogeneity of PRLR signaling dynamics affects downstream signaling outcomes. Our data underscore the versatility of building a generalizable modeling-imaging framework to quantitatively understand signal transduction in and beyond β-cells.

08

Jackson Chin

Abstract: Pneumonia, a respiratory infection leading to a build-up of fluid or pus in the alveoli, is the leading cause of death among infectious diseases. Despite its disease burden, the immunological signatures underpinning pneumonia mortality are poorly understood, possibly due to limited integrative studies examining immunological responses in infected tissues. Single-cell RNA-seq (scRNA) measurements collected from bronchoalveolar lavages (BAL) may provide high-resolution insights into localized immune responses in pneumonia, but the high-dimensionality of these datasets complicates effective analysis as measurements can exhibit variation across genes, localized to certain cell populations, with heterogeneous presentation between patients. Tensors are organized arrays of data, akin to high-dimensional versions of matrices, that may help to analyze such scRNA datasets. Each biological dimension (patients, genes, and cells) can be represented as dimensions in the tensor, preserving the inter-dimensional relationships and single-cell heterogeneity in scRNA measurements that traditional pseudo-bulk methods might miss. Tensor decomposition methods, like PARAFAC2, then reduce such tensor-structured data into interpretable components that capture immunological patterns observed across dimensions. Here, we apply PARAFAC2 to scRNA measurements collected from BAL samples from patients with severe pneumonia. PARAFAC2 successfully captures mortality-related patterns, predicting pneumonia mortality with an accuracy exceeding 70% and finding that immunological pathways underpinning mortality differ between patients with and without COVID-19 pneumonia. Interpretation of these pathways identifies mortality determinants at cellular resolution, finding that bacterial pneumonia mortality is driven by B cell dysfunction whereas macrophage-mediated granulocyte responses increase mortality risk in COVID-19 pneumonia. Subsequent integration of clinical characteristics reveals patient-driven variation in these signatures, finding that such B cell dysfunction is exclusive to immunocompromised patients while macrophage-mediated granulocyte responses are a product of extended infection and T cell exhaustion. Collectively, these efforts provide high-resolution insights into pneumonia mortality and identify new therapeutic targets for disease management.

09

Christine Cho

Abstract: Background: Pancreatic development, especially of the endocrine and exocrine compartments, is sensitive to early environmental perturbations. Peroxisome proliferator-activated receptors (PPARs) are nuclear receptors involved in metabolic regulation and development, with implications in diabetes pathogenesis. Zebrafish are the ideal model for tracking the development of the pancreas in real time with the use of microscopy and transgenic lines.

Objective: To investigate how modulation of PPAR signaling influences pancreatic development and morphology, and to identify potential pathways leading to early diabetic phenotypes.

Methods: Transgenic zebrafish expressing insulin:GFP and ptf1a:GFP were exposed to agonists and antagonists of PPARα, PPARβ/δ, and PPARγ, totalling 7 exposure groups including control, from 1 to 6 days post fertilization (dpf). Morphological assessments of the pancreas and whole embryos were conducted, alongside gene expression profiling.

Data Analysis: A multilayer network model was constructed, consisting of a network for each developmental endpoint. Morphology and gene expression data across time points. The multilayer networks were embedded and then clustered using persistent homology to identify endpoints associated with exposure conditions.

Conclusion: This study aims to characterize the temporal effects of PPAR modulation on pancreas development, offering insights into early molecular events in diabetic pathogenesis.

10

Nikki Chun

Abstract: A gene regulatory network (GRN) is a set of transcription factors that regulate the expression of genes encoding other transcription factors. The dynamics of GRNs explain how gene expression changes over time. GRNmap is a MATLAB software package that uses ordinary differential equations to model dynamics of small-scale GRNs. The program estimates production rates, expression thresholds, and regulatory weights for each transcription factor in the network based on DNA microarray data and then performs forward simulations of model dynamics. While the model has been successfully used to understand networks of 15-20 genes, we wanted to closely examine how it works on a smaller scale to determine parameter sensitivity. All 21 possible “toy” networks of 3 nodes and 4 edges were created based on the simulated expression data outputted when known parameters were run through a forward simulation. Then the simulated data was used to estimate the parameters again. Comparison of the known to estimated parameters showed that estimating production rates in addition to weights and thresholds reduced the accuracy of the results. The model was also sensitive to the direction and magnitude of the weight parameters for a network with the same connectivity. This finding led to a narrower analysis of just one of the network motifs: the feed forward loop, which has 3 nodes and 3 edges where the first transcription factor regulates both the second and third transcription factors and the second also regulates the third. Feedforward loops are classified as coherent or incoherent based on the regulatory inputs to the third transcription factor. Out of the 8 permutations of this one network motif, the ones that are found most common in nature (Types I and II) perform the best in our model. These results will help us to interpret the findings of the larger 15-20 gene models.

11

John Corrette

Abstract: The 3d organization of the genome — in particular, which two regions of DNA are in contact with each other— plays a role in regulating gene expression. Several factors influence genome 3d organization. Nucleosomes (where ~100 basepairs of DNA wrap around histone proteins) bend, twist and compactify chromosomal DNA, altering its polymer mechanics. How much does the positioning of nucleosomes between gene loci influence contacts between those gene loci? And, to what extent are polymer mechanics responsible for this? To address this question, we combine a stochastic polymer mechanics model of chromosomal DNA including twists and wrapping induced by nucleosomes with two data-driven pipelines. The first estimates nucleosome positioning from ATAC-seq data in regions of high accessibility. Most of the genome is low-accessibility, so we combine this with a novel image analysis method that estimates the distribution of nucleosome spacing from electron microscopy data. There are no free parameters in the biophysical model. We apply this method to IL6, IL15, CXCL9, and CXCL10, inflammatory marker genes in macrophages, before and after inflammatory stimulation, and compare the predictions with contacts measured by conformation capture experiments (4C-seq). We find that within a 500 kilo-basepairs genomic region, polymer mechanics with nucleosomes can explain 71% of close contacts. These results suggest that, while genome contacts on 100kbp-scales are multifactorial, they may be amenable to mechanistic, physical explanation. Our work also highlights the role of nucleosomes, not just at the loci of interest, but between them, and not just the total number of nucleosomes, but their specific placement. The method generalizes to other genes, and can be used to address whether a contact is under active regulation by the cell (e.g., a macrophage during inflammatory stimulation).

12

Chenlu Di

Abstract: While annotations of functional non-coding regions in the human genome are becoming more precise, the fitness effects of mutations in these regions remain unclear. We developed a pipeline to infer the distribution of fitness effects (DFE) of non-coding mutations using polymorphism data from the 1000 Genomes Project. Our analysis of enhancers, promoters, and conserved non-coding regions revealed distinct selection patterns. While mutations in enhancers are often neutral, approximately 30% of mutations in promoters are deleterious. The most conserved non-coding regions, showing reduced divergence across mammals and primates, have the highest proportion of deleterious mutations. However, primate-specific conserved sites have a higher proportion of deleterious mutations than regions that are conserved across mammals, but less conserved in primates specifically. Notably, we infer the most conserved sites across mammals, frequently targeted in biomedical studies, account for only a minority of deleterious mutations in noncoding regions. For example, the top 5% of conserved non-coding sites encompass fewer than 20% of deleterious mutations, indicating that functional non-coding regions are widely distributed across evolutionary constraints. Our findings highlight the dynamic evolution of gene regulation and shifting selection pressures over deep evolutionary timescales. These insights have broad implications for using comparative genomics to identify non-neutrally evolving sequences in the human genome.

13

Ethan Dickson

Abstract: The integrated Stress Response (ISR) acts as a central node for translational regulation to homeostatic attacks. In response to various stresses, one of four stress kinases, PERK, GCN2, HRI, and PKR phosphorylate eIF2α, shutting off global protein translation with select upregulation of ATF4 (adaptive) and CHOP (apoptotic). The ISR’s capacity to tightly regulate protein expression provides cancer with a node to tune translation to the needs of the tumor, but excessive activation can induce apoptosis, creating a threshold-dependent survival mechanism. Indeed, some cancers silence their ISR, allowing for high protein expression and cell proliferation, rendering them sensitive to ISR induced apoptosis. However, others exhibit constitutively activated ISR, leveraging ATF4 to enhance metabolic adaptation thereby activating factors that resupplying the tumor microenvironment and confer resistance to common chemotherapeutics.

Given the ISR acts as a central node of cancer robustness and apoptosis depending on the cancer and several common cancer treatments are ISR activators, there is a critical need to understand the nature of the ISR pathway in specific cancers to guide effective treatment strategies.

To address this, I have used Opto-PKR, an in-house optogenetic kinase engineered to phosphorylate eIF2α in the presence of 450 nm light, enabling specific ISR activation without off-target stress response activation. I have expressed opto-PKR in the NCI-60 cancer panel, in addition to select pancreatic and liver cancers. Using Immunofluorescence, I have measured each cancer’s adaptive ISR potential via ATF4 expression and apoptotic ISR potential via CHOP expression. 

Moving forward, these data will be analyzed using machine learning techniques like symbolic regression to derive quantitatively distinct ISR profiles. Finally, I will perform high-throughput viability assays combining ISR-associated cancer drugs with ISR potentiators and repressors to determine the relationship between a cancer’s ISR profile and its most effective treatment cocktail.

14

Elizabeth Elton

Abstract: Cancer-associated fibroblasts (CAFs) are an important element of the tumor microenvironment that play a key role in cancer metabolism. CAFs have also been shown to contribute to drug resistance in colorectal cancer (CRC). In order to understand the impact of CAFs on CRC metabolism, the Finley research group previously developed a computational network model of central carbon metabolism that integrates experimental data to represent KRAS-mutant and wild-type CRC cells cultured in CRC media or CAF-conditioned media. This model uses a systems biology approach to simulate the effects of enzyme knockdowns and analyze their system-wide impact on cancer cell metabolism. We also used machine learning methods to perform dimensionality reduction on these knockdown results to identify specific, impactful enzyme knockdowns. The results of this analysis revealed multiple impactful knockdowns that influence CRC cells. In this work, I further explored the metabolic perturbations shown to impact CRC cell metabolism using a flux sampling approach. Flux sampling methods are useful tools to understand the distributions of reaction fluxes. I applied flux sampling methods to the metabolic models representing KRAS-mutant cells cultured in CAF-conditioned media with and without enzyme perturbations. Such an analysis enables us to explore the feasible flux solution spaces. This analysis provides insight into the driving forces behind variability in metabolic outputs. Specifically, the flux sampling results highlight key differences between our base, unperturbed model and subsequent enzyme knockdown models. Overall, this work improves our understanding of the impact of CAF-mediated crosstalk on the metabolic landscape of CRC cells.

15

David Frankhouser

Abstract: Chronic myeloid leukemia (CML) is a myeloproliferative neoplasm defined by expression of the fusion genes BCR::ABL and its potential for disease transformation, yet the precise mechanisms and timing of its evolution remain poorly understood. Previously, we used time-series bulk transcriptome of peripheral blood mononuclear cells from a mouse model to construct a potential landscape that predicted CML development and progression via state-transition modeling. Here, we report on a longitudinal single cell (sc) RNA-sequencing in chronic phase (CP) and blast crisis (BC) CML mouse models to study the state-transitions at the single cell (sc) level. Different from the bulk transcriptome, we could not identify a disease state encoded in the sc-transcriptomes, as the cells were observed in a continuum of microstates occupying both health and leukemia states. However, combining the single cell transcriptomes into cell type pseudobulk (PsB) transcriptome samples we recapitulated a smooth transition from a health state to a CML state, and recapitulated that observed in our previous bulk transcriptome study. The PsB state-space was characterized by a three well potential landscape with three stable critical points defining the state-transition from health to leukemia. Importantly, by computing the PsB transcriptome for individual cell types, we first showed that the system level dynamics could be modeled as a linear combination of for individual cell types PsB transcriptomes of individual blood cell types. We validated this approach and our findings in a second independent time-series blast crisis (BC) mouse experiment. In conclusion, we offer a mathematical explanation for why phenotypes, such as health of leukemia, are not encoded in sc transcriptomes, which are microstates, but rather at the macrostate level. This analytical approach is likely applicable to other types of leukemia and cancer to predict disease evolution and treatment response.

16

Sebastian Griego

Abstract: Despite advances in antiretroviral therapy (ART), HIV eradication remains challenging due to persistent latently infected CD4+ T-cells. A hybrid approach combining differential equations with Physics-Informed Neural Networks (PINNs) is presented to predict latent reservoir dynamics in patients undergoing ART. Using clinical data from 28 acutely-infected patients, a viral dynamics model was developed incorporating interactions between uninfected CD4+ T-cells, infected cells, latently infected cells, and free virus under ART. This framework addresses the challenge of limited latency data while ensuring biologically plausible predictions through physics-constrained learning. The approach offers potential applications in personalizing treatment strategies and evaluating interventions targeting HIV reservoir reduction, with broad applicability to other biomedical scenarios with similar data constraints.

17

Xiaolu Guo

Abstract: Macrophages initiate pathogen-appropriate immune responses with the activation dynamics of transcription factor NFκB mediating specificity. Live-cell imaging revealed the stimulus-response specificity of NFκB dynamics among heterogeneous populations of cells. To study stimulus-response specificity beyond what is experimentally accessible, we develop mathematical model simulations that capture the cellular heterogeneity of stimulus-responsive NFκB dynamics and the stimulus-response specificity performance of the population. Complementing experimental data, extended-dose response simulations improved channel capacity estimates. By collapsing parameter distributions, we locate information loss to receptor modules, while the negative-feedback-containing core module show remarkable signaling fidelity. Further, constructing single-cell network models reveal the stimulus-response specificity of single cells. We find that despite stimulus-response specificity limitations at the population level, the majority of single cells are capable of responding specifically to immune threats, and that the few instances of stimulus-pair confusion are highly uncorrelated. The diversity of blindspots enable small consortia of macrophages to achieve perfect stimulus distinction.

18

Arya Hamidzad

Abstract: Methotrexate, while effective in managing rheumatoid arthritis (RA), can lead to the development of rheumatoid nodules in 5-10% of patients. Rheumatoid nodules are a common extra-articular manifestation of RA, occurring in approximately 20% of RA patients. These nodules most commonly form in areas subjected to repeated pressure, such as the elbows and finger joints, but can occasionally involve internal organs, rarely including the larynx. Laryngeal involvement of RA can lead to various symptoms, including hoarseness, pain, and airway obstruction. In severe cases, emergency interventions such as tracheostomy may be required to maintain a patent airway. We describe the case of a 62-year-old woman with a history of long-standing RA, presenting with dysphagia, mild biphasic stridor, and severe nodulosis. Over the past 22 years, she had been treated with methotrexate. A computed tomography (CT) scan of the neck revealed a 1.8 cm x 2.3 cm mass located behind the superior aspect of the right lateral thyroid cartilage. Direct laryngoscopy and biopsy confirmed the mass to be a rheumatoid nodule. Treatment with colchicine improved her dysphagia, and she ultimately underwent surgical partial resection to better alleviate her symptoms. This case highlights the rare but clinically significant potential for methotrexate-induced rheumatoid nodules affecting the larynx, highlighting the importance of early recognition to prevent airway compromise and improve patient care.

19

Timothy Hamilton

Abstract: The advent of single cell methods has revolutionized the fields of developmental and systems biology by enabling the investigation of individual cells in order to understand how their behavior contributes to the complex phenomenon observed at the tissue and organ level. Supporting these advancements is the fundamental idea that the gene expression of single cells, as measured by mRNA, once “denoised” and suitably transformed, can reveal distinct expression profiles that correspond to biologically distinct and meaningful groups. Unfortunately, recent work belies that notion; if a small percentage of cells is removed from the dataset and the typical analysis pipeline is repeated, the partitioning of remaining cells is significantly affected, with cells that previously existed in separate clusters being put in the same partition and cells previously in the same cluster split apart into different clusters. While a significant portion of this variability can be attributed to the community detection algorithms used to cluster scRNA-seq data, less work has been done to characterize whether the geometric structure of single cell rna-seq data fits with the current paradigm of biologically distinct, and statistically reproducible cell types. To answer this question, we developed a new method to determine the relative orientation of scRNA-seq data before and after the cells are removed from a dataset. Our findings imply that the analysis pipeline used to analyze scRNA-seq data are extremely sensitive to dataset composition, implying that the insights generated by the analysis of single cell data may not be generalizable beyond the original experiment.

20

Lun Hao

Abstract: The integrated stress response (ISR) is a conserved cellular signaling pathway that responds to various stress signals by modulating protein synthesis and restoring cellular homeostasis. Cancer cells exhibit remarkable plasticity, enabling them to adapt and survive under a broad range of stressors and the ISR can promote either protective or oncogenic outcomes, depending on the context. Despite its importance, studying the ISR has been hampered by the lack of tools that allow precise perturbation of stress signaling pathways without causing real damage. To address this, we developed opto-PKR, a novel optogenetic tool that allows dynamic, tunable, and reversible control of the ISR using light. This approach provides “pure,” on-target activation of the stress response, enabling high-resolution dissection of ISR dynamics.

We hypothesize that ISR re-wiring varies by cancer type and pathology, and that exploiting these differences could selectively kill cancer cells. To test this, we employed opto-PKR in the NCI-60 human tumor cell line panel using a custom-built 96-well Optoplate for precise light delivery. After light-induced ISR activation, we tracked the flow of stress signaling information through key nodes using immunofluorescence staining of p-eIF2α (phosphorylated eukaryotic initiation factor 2 alpha, a marker of ISR activation), ATF4 (a transcription factor associated with adaptive stress responses), CHOP (C/EBP homologous protein, linked to apoptosis), and G3BP (GTPase-activating protein-binding protein, a marker for stress granules). To quantify the high-dimensional data across cell lines, we developed an AI-based image analysis pipeline leveraging deep learning to segment cellular compartments for quantification. Preliminary data show that opto-PKR efficiently activated these signaling nodes across multiple lines, with unique dynamic responses observed between cancer types. Analysis showed distinct ISR sensitivities across cancer cells, with colon and lung adenocarcinoma cells displaying heightened sensitivity to stress-induced apoptosis marked by early expression of CHOP. In contrast, glioblastoma and astrocytoma cells required prolonged ISR activation to reach apoptosis and exhibited rapid downregulation of p-eIF2α, a central ISR signaling protein, suggesting a strong negative feedback mechanism that enables evasion of apoptosis. Additionally, we also employed ISR-specific modulators to assess their impact in conjunction with chemotherapy on enhancing the sensitivity of various cancer cells to stress-induced apoptosis, as evaluated by cell viability experiments.

These findings reveal distinct ISR wiring patterns in different cancers and underscore the potential of opto-PKR as a tool for dissecting cancer-specific stress mechanisms. Ongoing work aims to explore ODE models to integrate our dynamic single-cell data with ordinary differential equation (ODE) models to uncover mechanistic parameters, such as feedback strength or transcriptional delay, that govern ISR sensitivity. By combining model-based inference with dimensionality reduction and clustering techniques, we hope to identify core phenotypic signatures that explain why some cancer cells resist stress-induced apoptosis while others succumb. This approach may ultimately guide the rational design of combination therapies targeting ISR vulnerabilities in a cancer-type-specific manner.

21

Zachary Hemminger

Abstract: Genetic variation can alter brain structure and, consequently, function. Comparative statistical analysis of mouse brains across genetic backgrounds requires spatial, single-cell, atlas-scale data, in replicates—a challenge for current technologies. We introduce Atlas-scale Transcriptome Localization using Aggregate Signatures (ATLAS), a scalable tissue mapping method. ATLAS learns transcriptional signatures from scRNAseq data, encodes them in situ with tens of thousands of oligonucleotide probes, and decodes them to infer cell types and imputed transcriptomes. We validated ATLAS by comparing its cell type inferences with direct MERFISH measurements of marker genes and quantitative comparisons to four other technologies. Using ATLAS, we mapped the central brains of five male and five female C57BL/6J (B6) mice and five male BTBR T+ tf/J (BTBR) mice, an idiopathic model of autism, collectively profiling over 40 million cells across over 400 coronal sections. Our analysis revealed over 40 significant differences in cell type distributions and identified 16 regional composition changes across male-female and B6-BTBR comparisons. ATLAS thus enables systematic comparative studies, facilitating organ-level structure-function analysis of disease models.

22

Xinyi Hu

Abstract: This research investigates the impact of environmental temperature on the transmission dynamics of co-circulating high and low pathogenic avian influenza (AIV) strains in a non-migratory aquatic bird species. By extending a compartmental epidemiological model to include environmental viral load and temperature-dependent viral decay, we explore how seasonal fluctuations influence disease spread. The model distinguishes between direct bird-to-bird transmission and indirect fecal-oral transmission via contaminated water sources. Analytical and numerical results demonstrate that warmer temperatures significantly accelerate viral decay, thereby reducing environmental transmission. However, the reproduction number of highly pathogenic AIV (HPAI) can surpass that of low pathogenic AIV (LPAI) in warmer months, indicating a potential for seasonal surges. Sensitivity analyses and heatmaps reveal the critical role of indirect transmission and viral persistence in shaping outbreak dynamics. The findings highlight the importance of climate variability in predicting and managing avian influenza outbreaks in wildlife populations.

23

Helen Huang

Abstract: In response to infection or vaccination, lymph nodes must select antigen-reactive B-cells while eliminating auto-reactive B-cells. B-cells are instructed via B-cell receptor (BCR), which binds antigen, and CD40 receptor by antigen-recognizing T-cells. How BCR and CD40 signaling are integrated quantitatively to jointly determine B-cell fate decisions remains unclear. Here, we developed a differential-equations-based model of BCR and CD40 signaling networks activating NFκB. The model recapitulates NFκB dynamics upon BCR and CD40 stimulation, and when linked to established cell decision models of cell cycle and survival control, the resulting cell population dynamics. However, upon costimulation, NFκB dynamics were correctly predicted but the predicted potentiated population expansion was not observed experimentally. We found that this discrepancy was due to BCR-induced caspase activity that may trigger apoptosis in founder cells, unless timely NFκB-induced survival gene expression protects them. Iterative model predictions and sequential co-stimulation experiments revealed how complex non-monotonic integration of BCR and CD40 signals controls positive and negative selection of B-cells. Our work suggests a temporal proof-reading mechanism for regulating the stringency of B-cell selection during antibody responses.

24

Holly Huber

Abstract: Motivation. Computational models in biology are often underdetermined—that is, there is little data relative to the complexity/size of the model. This is primarily due to limits in our ability to observe specific biological systems and restricts the utility of these models. However, there are a growing number of experimental databases in biology. While these databases provide more observations, they often do not have observations that match the system of interest exactly. Here, we investigate what information can be gleaned from these large but general databases in the context of modeling a specific system. Ultimately, our goal is to better determine models of specific systems, thereby increasing their utility. 

Results. We use a Bayesian framework to quantify what information is gained from incorporating measurements from the protein data bank (PDB) and UniProt (FASTA) when modeling cell signaling. We choose to investigate these databases utility in the context of dynamic cell signaling models because experimental measurements of the variables of interest—protein dynamics—are still quite limited. Thus, the potential upside of incorporating general measurements from these databases should be higher. We find that we can successfully integrate measurements of sequence and structure from these databases by taking advantage of existing machine learning models and the unique composition of these signaling models. Integrating measurements from these databases improves parameter estimation of the cell signaling models. However, the impact of these measurements on predictions proves inconsistent, depending upon the complex relationship between model prediction and parameter values. In conclusion, this study demonstrates that large but general measurements from databases have the potential to better determine a subset of parameters in mechanistic models of cell signaling.

25

Rose Johnson-Leiva

Abstract: Hair follicles (HFs) are mini-organs in skin who undergo cyclic growth. During the anagen phase, the hair shaft is produced from the bottom part of a HF, referred to as the hair bulb. Proper regulations of the HF bulb cell fate decisions are crucial to maintain an anagen HF, therefore guaranteeing the continuous production of hair. Recent experiments have provided evidence on how the HF bulb is replenished during anagen. In this paper, we develop a hybrid multiscale computational model on HF bulb, integrating an agent-based submodel of cell divisions and movement, and a reaction-advection-diffusion equation submodel of diffusive signaling dynamics. Using our model, we investigate the HF bulb replenishment dynamics driven by different cell dividing strategies, showing that signaling-driven cell division may lead to efficient replenishment dynamics.

26

Vaibhava Kesarwani

Abstract: Effective vaccination requires the selection of B-cells with high affinity and specificity towards the immunization antigen. B-cell selection depends on the coordination of two signals, the first from the antigen receptor (BCR receptor) and the second from helper T-cells (CD40 receptor) which are separated by time. Prior work from our lab shows that effective selection depends on receiving the second signal within a certain time window of receiving the first, a concept we term temporal proofreading. Here, I explored whether there is an optional time window between the sequential stimulation of the two receptors that maximizes selection efficiency and B-cell population expansion, and what aspects of signaling in B-cells may control this. I investigated B-cell population dynamics under combinations of the two signals, using dye dilution assays in an vitro co-culture system that mimics physiological conditions within the germinal center in lymph nodes where vaccine responses take place. I discovered that varying dose combinations of BCR and CD40 stimuli produce different effects for B-cell fate outcome. Low signal conditions synergistically increase B-cell population expansion, whereas high signal conditions antagonistically diminish it. Additionally, I found 6 hours to be the optimal delay time between the two receptors for maximizing B-cell survival under most signaling conditions. I further tested the effect of pro-proliferative cytokines such as IL-6 and found that it enhanced cell survival but showed no differences for varying time delays. My results begin to reveal the molecular control of B-cell selection during a vaccine response. This can help design vaccine adjuvants that modulate selection and expansion of B-cell clones to maximize antibody mediated immunity while minimizing auto reactive adverse effects.

27

Jeffrey Kok

Abstract: Soybean oil (SO) consumption has increased dramatically over the past few decades, now accounting for over 60% of edible oil use in the U.S. While historically considered heart-healthy due to its high polyunsaturated fat content, recent research has revealed that SO, particularly through its high linoleic acid (LA) content, promotes obesity, insulin resistance, and nonalcoholic fatty liver disease (NAFLD) in mice. Contributing, possibly, to the rising prevalence of obesity and type 2 diabetes. Metabolomic and proteomic analyses have also implicated bioactive lipid metabolites known as oxylipins, derived from omega-6 fatty acid linoleic acid in soybean oil via the cytochrome P450 / soluble epoxide hydrolase (CYP/sEH) pathway, in mediating these effects. Yet it remains unclear whether these adverse effects are reversible.

In this study, we aim to investigate whether the metabolic consequences of a soybean oil-based high-fat diet (SO-HFD) can be reversed by either returning to a standard chow diet or implementing an every-other-day fasting (EODF) regimen. Additionally, we will examine whether sex differences influence susceptibility to metabolic disorders. Male and female C57BL6/N mice were assigned to 6 groups: continuous SO-HFD, standard vivarium chow (control), standard reversal (SO-HFD until 40gm followed by 5 weeks chow), EODF vivarium-reversal (SO-HFD until 40gm followed by 5 weeks chow w/ EODF), EODF SO-reversal(SO-HFD until 40gm followed by 5 weeks SO-HFD w/ EODF), or short-term reversal (4 weeks SO-HFD followed by 8 weeks chow). Changes in body composition, glucose, and ketone body levels, and gene expression were analyzed.

We hypothesize that early dietary reversal will restore metabolic health, while longer SO-HFD exposure may cause persistent alterations mitigated by EODF fasting. This study will provide novel insight into the reversibility of diet-induced metabolic damage and the differential effects across sexes, informing both nutritional guidelines and therapeutic strategies for metabolic disease.

28

Leonila Lagunes

Abstract: Cells employ large macromolecular machines for the execution and regulation of many vital processes for cell and organismal viability. Interestingly, cells cannot synthesize these machines as functioning units. Instead, cells synthesize the molecular parts that must then assemble into the functional complex. An extremely common motif is a stacked ring-like topology. Thus, understanding how stacked rings assemble is crucial for our understanding of how macromolecular complexes are regulated. In this work, we developed a mathematical deterministic model of stacked trimer assembly that accounts for different binding affinities between and within rings. Our main finding is that deadlock – a severe form of kinetic trapping– can be extremely long, lasting for days or longer. Deadlock is worst when all the interfaces have high binding affinities. Therefore, we predict that evolutionary pressures select against stacked trimers having strong binding affinities throughout. We tested our prediction by analyzing solved stacked trimer structures; we found that indeed the majority – if not all – of the stacked trimers did not contain very strong interactions. Finally, to better understand the origins of deadlock, our pathway analysis show that when all the binding affinities are strong, many of the possible pathways are utilized, consuming subunits, and creating high levels of deadlock. In sum, our work provides critical insight into the evolutionary pressures that have shaped the assembly of stacked rings.

29

James Le

Abstract: Cell polarity is essential for the organization and function of tissues. In the Drosophila wing disc, the Fat-Dachsous (Ft-Ds) signaling pathway manages planar cell polarity (PCP) via the heterodimer formation which leads to asymmetric distribution across the membranes of adjacent cells. To better understand this process, we study and develop a mathematical model that incorporates Ft-Ds biochemical interactions and binding, different localized feedback, including self-promotion and mutual inhibition, and intercellular coupling. Through bifurcation analysis and numerical simulations, we identify conditions under which the system exhibits bistability, leading to robust polarization at the cellular level. Significantly, our results show that localized feedback and asymmetric Ds (Ft) distribution contribute to the formation of stable tissue-wide polarization patterns and the bistability under a range of conditions. Our work provides new insights into the cellular mechanisms underlying robust polarity in epithelial tissues and highlights the importance of feedback regulation in polarity formation.

30

John Leano

Abstract: John Leano, Gayatri Raut, Suhasini Guttalu, Rachel Jones, Jeffrey Kok, Jose Martinez-Lomeli, Gary Chen, Poonamjot Deol, Frances M Sladek Department of Molecular, Cell and Systems Biology, University of California Riverside Hepatocyte Nuclear Factor 4alpha (HNF4a) is a master regulator of liver-specific gene expression, regulating both carbohydrate and lipid metabolism. The HNF4A gene has two promoters (P1 and P2) that drive the expression of multiple transcript variants – P1 which is expressed in normal adult livers, while P2 is expressed in fetal livers and under various conditions of stress in adults. In this study, we employ HNF4a exon swap mice that express either P1- or P2-HNF4a in a systems biology approach to examine the role of the HNF4a isoforms under conditions of metabolic stress induced by fasting, cold exposure, exercise and aging. The results suggest that the HNF4a isoforms play a role in mitochondrial function in the liver as well as other tissues, such as the heart and brown adipose tissue (BAT). Furthermore, HNF4a’s role appears to be impacted by the sex of the mice.

31

Timmy Liang

Abstract: HIV is a global issue with approximately 27 million people with HIV+ reported by CDC 2019. Of the total infected population, over 30% are related to drug-abuses. To control HIV infection, tremendous efforts are applied to prevent the spread of HIV through antiretroviral therapy (ART), however, there is no cure for HIV especially due to the establishment of latently infected cells. In addition, the contributing factors from drugs of abuse, such as opiates or fentanyl, have been shown to increase the viral load, and how drugs of abuse affect the HIV latent infection is not well-understood. In this study, we develop a mathematical model to investigate the pharmacodynamic efficacy of the drugs of abuse and how they can affect latent infection dynamics. The model is validated using experimental data from HIV infection of humans and Simian Immunodeficiency Virus (SIV) infection of morphine-addicted macaques, and using parameters that are obtained via research or studied from published articles. Our model shows that the dynamics of latently infected cells can be significantly altered due to the presence of the drugs of abuse, while accounting for the logistic drug degradation factor with time.

32

Diamond Mangrum

Abstract: Treatments such as chemotherapy, radiation, and immunotherapy all harness the critical role of apoptosis in maintaining healthy tissue function by balancing cell proliferation and cell death. This delicate equilibrium removes damaged or unwanted cells, preventing unchecked growth and preserving tissue integrity. In contrast to necrosis, which leads to the uncontrolled release of inflammatory substances, apoptosis provides a more controlled and immune-friendly response. By modulating both intrinsic and extrinsic signaling pathways, apoptosis can be precisely regulated in therapeutic interventions, minimizing collateral damage to healthy tissue. However, in cancers driven by hormonal signaling, the apoptotic system becomes dysfunctional. The model, expanded from previous work, applies quantitative analysis tools to simulate the effects of three common prostate cancer therapies, aiming to identify synergistic approaches that enhance therapeutic efficacy and promote apoptosis in prostate cancer cells.

33

David Mastro

Abstract: B-Lymphoblastic Leukemia (B-ALL) is a cancer in which malignant B-lymphoblasts clonally expand in the bone marrow. Children and adolescents are often affected, with frequent recurrence after initial treatment. In other leukemias (Acute Myeloid Leukemia, AML), relapsed disease is partially attributed to Leukemic Stem Cells (LSCs), a rare population of quiescent malignant cells that can evade therapy and replenish the leukemic clone. However, whether LSCs contribute to the high relapse rate in B-ALL (~15% of pediatric cases) remains unknown. Thus, we asked whether putative LSCs could be identified in B-ALL single cell RNA sequencing (scRNA-seq) datasets. We applied an established AML LSC gene signature to recognize putative LSCs across multiple subtypes. A distinct metabolic and quiescent transcriptomic state was uncovered within these putative LSCs, suggesting LSCs may play a unique role in disease relapse in B-ALL. These results may inform therapeutic strategies to prevent relapse.

34

Ethan Nowaski

Abstract: One of the most important open questions in developmental biology is determining how organs and tissues form and maintain their shape. The robust formation of organs during development depends on the careful regulation of cellular processes such as cell adhesion, mechanical stiffness of the cell membrane, and internal pressure to create the tissue-scale architecture. A sophisticated communication system coordinates these developmental processes. This project utilizes the fruit fly wing imaginal disc, a powerful biological model system, to study how downstream effectors of ecdysone signaling contribute to regulating cell mechanical properties that influence cell shapes and overall tissue structures during wing disc eversion. Eversion is one of the final parts of development involving decline of cell proliferation and folding of the wing driven by cell rearrangements and morphological changes driven by actomyosin dynamics. The project workflow includes iteration between quantitative experiments and biologically calibrated computational model simulations. In this talk, a novel extension of the two dimensional subcellular element computational model of the developing Drosophila wing [1] will be described and model simulations will be used to demonstrate potential mechanisms of wing folding during eversion.
Nilay Kumar, Jennifer Rangel Ambriz, Kevin Tsai, Mayesha Sahir Mim, Marycruz Flores-Flores, Weitao Chen, Jeremiah J. Zartman and Mark Alber [2024], Balancing competing effects of tissue growth and cytoskeletal regulation during Drosophila wing disc development. Nat Commun 15, 2477 (2024).

35

Peace Olatoyinbo

Abstract: The SARS-CoV-2 spike protein contains variant-modified epitopes in both structured and intrinsically disordered regions (IDRs), such as the furin cleavage loop. It has been proposed that structural disorder might promote broader B cell recognition by accommodating sequence variation. To test this, we immunized mice with full-length Omicron BA.1 spike and profiled B cell specificity using single-cell BCR sequencing coupled with a panel of barcoded linear peptides spanning both structured RBD regions and disordered furin loop variants. While disordered peptides elicited robust responses, binding was strictly sequence-specific, with no tolerance for even minimal mutations across variants. Computational docking revealed that despite high sequence similarity among furin loop peptides, structural modeling and RMSD analysis showed significant conformational shifts between variants. Antibody–peptide binding was localized and clonotype-specific, with structural overlays confirming distinct engagement sites even within disordered regions. These findings challenge the assumption that IDRs promote cross-reactivity and instead highlight sequence identity as the key determinant of B cell recognition, even within structurally flexible epitopes.

36

Audrey Oliver

Abstract: Effective management for preventing and controlling of vector-borne infectious diseases depends on accurate estimates of the basic reproduction number (R0) and the effective reproduction number (Rt), which measure disease outbreaks and transmission trends. However, existing maximum likelihood-based methods often fail to capture the complexities of climate-influenced indirect transmission via vectors, potentially limiting their accuracy in complex epidemiological settings. In this study, we first develop a random forest-based method with climate predictors to map weekly dengue fever case data from Nepal to daily case data. We then create a data-driven modeling framework to accurately estimate R0 and Rt for climate-influenced vector-borne diseases. By leveraging the imputed daily case data, our framework provides improved estimates of reproduction numbers of climate-influenced dengue fever in Nepal. This approach allows us to better capture the dynamic interactions between vectors and humans that affect climate-influenced disease outbreaks and transmission patterns.

37

XINRU QIU

Abstract: Single-cell RNA sequencing analysis reveals distinct platelet subpopulations that may serve as critical biomarkers in COVID-19 and sepsis patient outcomes. Using an integrated approach combining machine learning, Universal Cell Embeddings (UCE), and single-cell transcriptomics, we identified three key platelet subpopulations: coagulation-active, hypoxic, and quiescent. The platelet-to-T-cell ratio emerged as a significant predictor of survival outcomes (AUC 0.754). Our analysis demonstrated that platelet heterogeneity and their interactions with other immune cells, particularly T-cells and endothelial cells, play crucial roles in disease progression. The study utilized data from multiple COVID-19, sepsis, and SLE patient cohorts, employing foundation models for robust cell type annotation and trajectory analysis. These findings suggest that early platelet precursor populations could serve as potential therapeutic targets, while distinct platelet signatures may help predict disease trajectories and inform treatment strategies.

38

Negin Rahimzadeh

Abstract: Alzheimer's Disease (AD) is a neurodegenerative disorder characterized by progressive cognitive decline, driven by complex molecular and cellular dysregulation across diverse brain cell types. To systematically interrogate these mechanisms, this project constructs a harmonized single-nucleus RNA sequencing (snRNA-seq) atlas by integrating data from 12 independent AD studies spanning control, mild cognitive impairment (MCI), and AD conditions. Using probabilistic reference mapping (scANVI), we generate a unified, cell-type-annotated atlas of the aging and diseased human brain. This PAN-AD Atlas forms the foundation for downstream analyses aimed at dissecting transcriptional alterations across disease states. Specifically, we perform conditional and sex-stratified differential gene expression analysis, identify cell-type-specific co-expression modules, characterize intercellular communication networks, and infer gene regulatory circuits. This integrative framework enables the discovery of conserved transcriptional programs and signaling pathways across disease states, advancing our understanding of the cellular mechanisms that drive AD pathogenesis.

39

Andrew Ramirez

Abstract: Effective tools for exploration and analysis are needed to extract insights from large-scale single-cell measurement data. However, current techniques for handling single-cell studies performed across experimental conditions (e.g., samples, perturbations, or patients) require restrictive assumptions, lack flexibility, or do not adequately deconvolute condition-to-condition variation from cell-to-cell variation. Here, we report that the tensor decomposition method PARAFAC2 (Pf2) enables the dimensionality reduction of single-cell data across conditions. We demonstrate these benefits for single-cell RNA-sequencing (scRNA-seq) experiments of peripheral immune cells in systemic lupus erythematosus (SLE) and healthy patient samples. By isolating relevant gene modules across cells and conditions, Pf2 enables straightforward associations of gene variation patterns across specific patients while connecting each coordinated change to certain cells without pre-defining cell types. Thus, Pf2 provides an intuitive universal dimensionality reduction approach for multi-sample single-cell studies across diverse biological contexts.

40

Mst Suraiya Akter Shefa

Abstract: Many types of biological cells, including cancer and endothelial cells, exhibit durotaxis, which is the motion in response to the elastic stress of the substrate. For example, durotaxis can drive cell migration along stiffness gradients and lead to non-trivial interactions between durotactic cells. While durotaxis is common among motile cells, the collective dynamics driven by durotaxis remain largely unexplored. In our work, we extended a recently developed computational PDE model for the collective motion of durotactic motile cells to explore the clustering dynamics of these cells and their dependence on key biophysical parameters. First, we demonstrate that an individual cell on a substrate with non-homogeneous bending stiffness prefers softer or stiffer regions, depending on the choice of modeling parameters. Next, we investigate the dynamics of cell doublets and show that two durotactic cells exhibit persistent rotations. Finally, in larger clusters (10–20 cells), the overall displacement is minimal, while cells inside the cluster continue to move actively. We also examine how the effective diffusion coefficient of such a cluster is determined by the properties of the individual cells within it. Overall, our findings provide new insights into the collective dynamics of active matter systems with elastic interactions.

41

Navaira Sherwani

Abstract: Understanding the biochemical and mechanical cues that cause epithelial sheets to reorganize themselves into three-dimensional organs remains a central challenge in biology. During Drosophila wing-disc eversion, a pseudostratified pouch folds outward and fuses into a bilayer that goes on to form the adult wing. This project unites quantitative experiments with a 3D multiscale mechano-chemical (M3D) model to decode the late-stage morphogenesis of the Drosophila wing imaginal disc during eversion. The work couples: (i) a GPU-accelerated particle-based model that resolves apical, basal and extracellular-matrix mechanics on a deforming epithelial surface; (ii) reaction–diffusion models for hormone (ecdysone) and morphogen signaling (Dpp, Wg) extended down to intracellular Rho1/Cdc42 dynamics; and (iii) machine-learning pipelines—Gaussian-process surrogate modeling, Bayesian optimization and neural-network solvers—to calibrate and accelerate simulations against time-lapse light-sheet imaging, biomechanical perturbations and quantitative immunostaining. Iterative experimentation will map how spatially patterned actomyosin contractility, cell-ECM adhesion and ECM stiffness drive coordinated cell reshaping, layer coupling and tissue folding. The resulting framework will yield predictive, systems-level insights into how hormonal timing interfaces with morphogen gradients to orchestrate organ-scale shape changes.

42

Luis Solano

Abstract: Advances in healthcare are accelerating the growth of an increasingly aging population on a global scale. Alzheimer’s disease (AD) is a debilitating neurodegenerative disorder, typically associated with aging and dementia. In AD, a system of cellular and molecular constituents contributes toward dysregulated cognition. Thus, the 90+ study aims to enroll and evaluate individuals over the age of 90 to be evaluated every 6 months until death, when single nuclei resolution gene expression of the donated brain will be measured. Having processed the first 100 samples, we begin our analysis with an ontology-level comparison of a group with severe dementia and high pathology juxtaposed to a group with low dementia and pathology. We report 2140, 5822, 5618, 5038, 1932, 316, 733, and 116 significantly (p.adj<0.05) differentially expressed genes in astrocytes, excitatory neurons, inhibitory neurons, microglia, oligodendrocytes, oligodendrocyte progenitors, vascular cells, and pericytes, respectively. Furthermore, we report 135, 20, 55, 23, 32, 36, 75, and 2 significantly enriched pathways (p.adj<0.05) for each cell type, respective

43

Niki Tavakoli

Abstract: Cancer-associated fibroblasts (CAFs) play a key role in metabolic reprogramming and are well-established contributors to drug resistance in colorectal cancer (CRC). To exploit this metabolic crosstalk, we integrated a systems biology approach that identified key metabolic targets in a data-driven method and validated them experimentally. This process involved high-throughput screening to investigate the effects of enzyme perturbations predicted by a computational model of CRC metabolism to understand system-wide effects efficiently. Our results highlighted hexokinase (HK) as one of the crucial targets, which subsequently became the focus of our experimental validation using patient-derived tumor organoids (PDTOs). Through metabolic imaging and viability assays, we found that PDTOs cultured in CAF conditioned media exhibited increased sensitivity to HK inhibition. Our approach emphasizes the critical role of integrating computational and experimental techniques in exploring and exploiting CRC-CAF crosstalk.

44

Mao Tian

Abstract: The exponential growth of AI in science necessitates efficient and scalable solutions for retrieving and preserving research information. Here, we present a tool for the development of a customized question-answer (QA) dataset, called Interactive Trained Research Innovator (iTRI) - QA, tailored for the needs of researchers leveraging language models (LMs) to retrieve scientific knowledge in a QA format. Our approach integrates curated QA datasets with a specialized research paper dataset to enhance responses' contextual relevance and accuracy using fine-tuned LM. The framework comprises four key steps: (1) the generation of high-quality and human-generated QA examples, (2) the creation of a structured research paper database, (3) the fine-tuning of LMs using domain-specific QA examples, and (4) the generation of QA dataset that align with user queries and the curated database. This pipeline provides a dynamic and domain-specific QA system that augments the utility of LMs in academic research that will be applied for future research LM deployment. We demonstrate the feasibility and scalability of our tool for streamlining knowledge retrieval in scientific contexts, paving the way for its integration into broader multi-disciplinary applications.

45

Talyssa Topacio

Abstract: The dust emitted from the desiccating Salton Sea and surrounding playa may be responsible for the corresponding incidence of chronic respiratory disease observed for inhabitants of the region. Our previous work showed that mice exposed to dust collected around the Salton Sea exhibited a “chronic-innate” pulmonary response characterized by neutrophil-dominant inflammation in host lung. We examined the microbiomes from these same lungs extracted from Salton Sea dust-exposed mice via targeted amplicon sequencing of 16S V3-V4 rRNA to investigate microbial community composition from dust-exposed mice as compared to ambient air exposure and likewise characterized host immune response. Our findings show that exposure to Salton Sea dust collected nearer to the sea significantly shifts the composition of the lung microbiome, while the microbiome of mice exposed to dust collected further from the sea more closely resembles that of mice exposed to dry, filtered air. Overall, we found that lung microbiome diversity and composition may shift independently from host pulmonary inflammatory response, suggesting a greater influence from environmental dust composition or microbial community interactions.

46

Angeliz Vargas Casillas

Abstract: Keloids are debilitating skin scarring disorders, triggered by an aberrant wound healing program and exhibit continuously spreading growth. Recent single cell RNA-sequencing experiments reveal heterogeneity in human keloid fibroblasts, and interactions among various sub-clusters of fibroblasts together with interactions with immune cells might be key factors regulating the keloid propagation. We developed an agent-based model to investigate the propagative dynamics of keloids, where single cell data inferred cell communications are implemented into the model. Using the model, we propose the potential regulating mechanisms behind keloid propagation.

47

Katie Vineall

Abstract: Behind mitochondrial function is phosphatase and tensin homologue-induced kinase 1 (PINK1), a serine/threonine kinase that plays roles in mitophagy, cell death, and regulation of cellular bioenergetics. Current approaches to studying PINK1 biology depend on bulk approaches that can only provide snapshots of activity, and can miss the dynamics and heterogeneity of PINK1 activity. Therefore, we sought to temporally characterize PINK1 activity by developing a novel PINK1 kinase activity reporter (KAR). In particular, KARs have been developed to characterize the activity of kinases such as protein kinase A, ERK, Akt, and many others at the single cell level. Towards developing a KAR for PINK1, I have constructed PINK1-SPARK (Separation of Phases-based Activity Reporter of Kinase), a novel tool for the temporal characterization of PINK1 activity based on phase separation. To develop PINK1-SPARK, I have inserted different PINK1 phosphorylation motifs and phosphoamino acid binding domains into the SPARK construct. With our candidate PINK1-SPARK, we observe the formation of puncta in U2OS cells treated with CCCP for 2 hours. No formation of puncta is observed upon treatment with dimethyl sulfoxide (DMSO), or when cells expressing PINK1-SPARK are treated with previously identified small molecule inhibitors of PINK1. Furthermore, we have validated the use of PINK1-SPARK in multiple cell types and have used PINK1-SPARK to characterize PINK1 activity following the addition of previously identified small molecule activators of PINK1. Thus, PINK1-SPARK provides a promising way to temporally characterize PINK1 activity in single cells, allowing for further elucidation of the role of PINK1 in mitophagy and cell function.

48

Khoi Vo

Abstract: Bacterial-fungal interactions play a fundamental role in many processes including crop biofuel development and biosystem design. In this work, we focus on the interactions between the fungus, Laccaria bicolor, and the bacterium, Psuedomonas fluorescens, and their integral role in the fitness of the roots of Populus species. Laccria bicolor synthesizes malate which stimulates growth and chemotaxis of P. flourescens. Furthermore, P. flourescens provides L. bicolor with thiamine thereby increasing fungal mass. We developed a multiscale computational model to investigate these interdependent interactions.The growth and branching of the fungal mycelia are modeled using an off-lattice spatial discrete submodel which is dependent on both diffusive and active translocation of internal nutrients and uptake of external nutrients. Malate secretion acts as a source of diffusive chemoattractant for P. fluorescens. The bacteria colony are represented by point sources of diffusing thiamine.

49

Charlotte Weymer

Abstract: This study aimed to characterize geographically diverse wild isolate strains of C. elegans, particularly CB4856, MY23, DL238, ECA36, JU2001, and XZ1516 in conjunction with the lab strain N2 and the transgenic strain JIM113. Label-free automated lineage tracing was achieved through the use of embGAN, a previously developed deep learning pipeline. The intersection branch distance was applied towards characterizing heterogeneity through cell cycle timing analysis. This metric is a measure of graph edit distance, modelling the invariant development as a weighted binary tree and assigning the weights with the cell cycle timing. Significant divergence in cell cycle timing was demonstrated by ECA36 and XZ1516 across whole-embryo cell cycle timing as well as in sublineage analysis with XZ1516 displaying high levels of heterozygosity. An additional Hawaiian strain CB4856 similarly displayed divergence due to two spontaneously developmentally delayed embryos. Clustering analysis revealed four clusters of embryos with many of these strains(N2, JIM113, CB4856 DL238, JU2001, and MY23) exhibiting identifiable yet relatively minimal levels of intra-strain heterogeneity. Future work will further characterize spatial heterogeneity throughout development through embryo alignment and application of the intersection branch distance.

50

Emerald Win

Abstract: The Drosophila wing disc is a model system which studies a tissue which eventually becomes the adult wing of a fruit fly. This model has primarily been used to study the underlying mechanisms for the development of growing epithelium; this is because of its short cell life, limited number of cells, and large number of shared genes amongst other mammals. "The relative simplicity and accessibility of the wing disc, combined with the wealth of genetic tools available in Drosophila, have combined to make it a premier system for identifying genes and deciphering systems that play crucial roles in animal development. Studies in wing imaginal discs have made key contributions to many areas of biology, including tissue patterning, signal transduction, growth control, regeneration, planar cell polarity, morphogenesis, and tissue mechanics." A large portion of cell development is dependent on the chemical regulation on the tissue. We consider a tissue-level morphogen by the name Dpp which is responsible for the regulation of cell growth and division through a downstream regulation of many other morphogens. Combining the receptor Tkv, the Dpp-Tkv complex ends up activating and inhibiting many other morphogens which, in turn, determines the regulation of two critical intracellular signals, Rho1 and Cdc42. In particular, Rho1 and Cdc42 together control actomyosin contractility, bridging together chemical and mechanical properties within the wing disc tissue. It is commonly agreed that studying the effects of morphogens when regulating mechanogens is critical to understanding proper tissue formation, that is why we aim to run experimental and mathematical models together to help map a deeper comprehension of the relationship between Dpp and Rho1,Cdc42.

51

Yue Wu

Abstract: Transient expression of the reprogramming factors Sox2, Oct4, Klf4 (OSK) has been shown to reverse epigenetic aging and rejuvenate metabolic performance. To screen for OSK-inducible anti-aging molecules in a high-throughput and efficient manner, we're developing cellular reporters to report on OSK activation.

52

Xin Xiang

Abstract: Understanding the mechanisms underlying cell polarization and morphogenesis is fundamental to developmental biology. This project proposes a multiscale mathematical framework to investigate the formation of interdigitated jigsaw-like patterns in pavement cells (PCs) of the Arabidopsis leaf epidermis. By integrating biochemical signaling, mechanical forces, and cell-to-cell interactions, the model seeks to elucidate how extracellular plant hormone auxin orchestrates local and global cell behaviors during tissue development.

53

Mark Xiang

Abstract: Vaccine responses depend on the Darwinian evolution of B-cells to generate high affinity antibodies. However, epigenetic variability produces cell fate decisions that appear stochastic, and then inherited within proliferative clonal bursts. How does the alternating fragility and stability of B-cell epigenetic states affect antibody generation? Integrating a wealth of immunological knowledge we computationally simulated antibody affinity maturation and validated the model with mutant mouse strains. We discovered that stochasticity in cell fate decisions does not impair but enables efficient affinity maturation. During clonal bursts epigenetic heritability further enhances genetic evolution by maintaining maximal proliferative fitness. Our work reconciles classical B-cell clonal selection theory with the epigenetic variability dynamics that are observed experimentally. Further, the interpretable knowledge-based modeling framework may assist the development of personalized vaccination.

54

Minxiao Yang

Abstract: Lung adenocarcinoma (LUAD) is the most prevalent subtype of lung cancer with a high degree of biological heterogeneity, including cell of origin, histological presentation, therapeutic response and clinical outcome, all of which pose challenges for precision diagnosis, prognosis and effective therapeutics. Our previous research in mouse models determined that Gram-domain-containing 2 (GRAMD2)-positive alveolar epithelial type 1 (AT1) cells, in addition to the well-recognized surfactant-secreting type 2 (AT2) cells, can serve as a LUAD cell of origin, presenting distinct histological, cellular, molecular features and microenvironmental landscape. However, the role of fibroblasts in influencing LUAD progression based on cell of origin remains unknown. To address this challenge, we leveraged the previously published snRNA-seq data to profile the transcriptomics of mesenchymal cells and observed that alveolar fibroblasts (AFs) in Sftpc+ AT2 cell-derived LUAD expressed higher levels of ECM-remodeling (Mmp2, S100a4), immunomodulatory (Cxcl14, Cxcl12), and metabolic genes (Fabp5), supporting the creation of a pro-metastatic niche. Pathway analysis revealed AFs from Sftpc+ AT2-derived LUAD exhibited significant enrichment in pathways associated with RNA biosynthetic processes and post-transcriptional regulation, whereas AFs from Gramd2+ AT1-derived LUAD showed pathways related to reprogrammed metabolism, ribosome biogenesis and ECM integrity. In addition, Sftpc+ AT2-derived LUAD cells harbored significantly higher chromosomal instability, as evidenced by elevated copy number variation (CNV) scores, relative to Gramd2+ AT1-derived LUADs. Differential gene expression and Gene Set Enrichment Analysis (GSEA) uncovered that Sftpc+ AT2-derived LUAD cells were enriched for pathways driving epithelial-mesenchymal transition (EMT), Wnt/β-catenin signaling, metabolic reprogramming, and RUNX2 transcriptional activity—factors closely associated with invasion, proliferation, and immune evasion. In contrast, Gramd2+ AT1-derived LUAD cells maintained epithelial integrity, junctional organization, and neurodevelopmental features, indicating a less invasive phenotype. Moreover, longitudinal PET/MRI imaging results suggested that Sftpc+ AT2-derived LUAD exhibited diffuse, heterogeneous, and metabolically active tumors, while Gramd2+ AT1-derived LUAD displayed localized, morphologically stable growth. Radiomic metrics confirmed persistent spatial heterogeneity in Sftpc+ AT2-derived LUAD, indicating elevated invasive and metastatic potential. Together, these findings highlight that the cell of origin shapes LUAD progression, with Sftpc+ AT2-derived LUAD exhibiting enhanced chromosomal instability, pro-metastatic fibroblast activation, and aggressive tumor phenotypes compared to their AT1-derived counterparts.

55

Stephanie Young

Abstract: A 3D chaos game is shown to be a useful way for encoding DNA sequences. Since matching subsequences in DNA converge in space in 3D chaos game encoding, a DNA sequence’s 3D chaos game representation can be used to compare DNA sequences without prior alignment and without truncating or padding any of the sequences. Two proposed methods inspired by shape-similarity comparison techniques show that this form of encoding can perform as well as alignment-based techniques for building phylogenetic trees. Simulations show that the proposed methods produce distances that reflect the number of mutation events; additionally, on average, distances resulting from deletion mutations are comparable to those produced by substitution mutations.

56

Guanqiao Yu

Abstract: High-grade astrocytoma with piloid features (HGAP) is a rare, newly classified tumor with ambiguous biological characteristics that challenge conventional treatment strategies. HGAP shares histological traits with both aggressive glioblastoma and lower-grade pilocytic astrocytoma, suggesting unique origins and behavior. The tumor’s rapid progression and limited survival outcomes makes it imperative to explore its underlying molecular mechanisms to improve diagnosis and therapeutic targeting. To study the molecular mechanisms of HGAP, we employed spatial transcriptomics on primary tumor specimens, classified based on their grading levels: higher-grade, lower-grade, and non-neoplastic areas. We then integrated our differentially expressed genes (DEGs) with publicly available datasets: The Cancer Genome Atlas (TCGA) for glioblastoma and Gene Expression Omnibus (GEO) for pilocytic astrocytoma (PA). This integrative approach enabled us to identify unique DEGs and pathways specific to HGAP that differentiates it from both GBM and PA. Subsequently, we intersected DEGs from HGAP’s higher-grade, lower-grade, and non-neoplastic areas, with a difference-in-difference analysis approach and identified the genes that undergo high-grade transformation.

57

Patrick Yuan

Abstract: The generation of high-affinity antibodies in germinal centers (GCs) requires B cells to integrate multiple external signals—namely, antigen uptake via the BCR and co-stimulation via CD40L from T follicular helper (Tfh) cells. This is a spatiotemporally complex process: antigens are non-uniformly distributed, B cells migrate stochastically, and the timing of signal acquisition can vary widely across cells. How this noisy, asynchronous environment gives rise to consistent and selective immune responses remains an open question.

Using imaging-derived data, we model antigen distribution as a thresholded fractional Brownian field and B cell migration as a persistent or Lévy walk. Within this dynamic extracellular context, we embed a coarse-grained ODE system that integrates temporally distinct signals from BCR and CD40 pathways, enabling us to investigate the consequences of signaling mismatches. Our simulations reproduce experimentally observed population dynamics across varying co-stimulation timings and reveal how tissue architecture, cell motility, and signal kinetics shape the stringency and efficiency of selection.

Our simulations replicate experimental observations of population dynamics under varied co-stimulation delays, and further reveal how spatial and temporal coordination enables the immune system to enforce stringent but flexible selection windows. These selection windows are essential for eliminating self-reactive clones while preserving high-affinity responders—underscoring the importance of temporal proofreading and spatial organization in maintaining both immune competence and self-tolerance.

58

Brian Zhang

Abstract: The manipulation of DNA replication and transcription can be harnessed to control cell fate. Central to the regulation of these DNA-templated processes are histone chaperones, which in turn are emerging as cell fate regulators. Histone chaperones are a group of proteins with diverse functions that are primarily involved in escorting histones to assemble nucleosomes and maintain the chromatin landscape. Whether distinct histone chaperone pathways control cell fate and whether they function using related mechanisms remain unclear. To address this, we performed a screen to assess the requirement of diverse histone chaperones in the self-renewal of hematopoietic stem and progenitor cells. Remarkably, all candidates were required to maintain cell fate to differing extents, with no clear correlation with their specific histone partners or DNA-templated process. Among all the histone chaperones, the loss of the transcription-coupled histone chaperone SPT6 most strongly promoted differentiation, even more than the major replication-coupled chromatin assembly factor complex CAF-1. To directly compare how DNA replication- and transcription-coupled histone chaperones maintain stem cell self-renewal, we generated an isogenic dualinducible system to perturb each pathway individually. We found that SPT6 and CAF-1 perturbations required cell division to induce differentiation but had distinct effects on cell cycle progression, chromatin accessibility, and lineage choice. CAF-1 depletion led to S-phase accumulation, increased heterochromatic accessibility (particularly at H3K27me3 sites), and aberrant multilineage gene expression. In contrast, SPT6 loss triggered cell cycle arrest, altered accessibility at promoter elements, and drove lineage-specific differentiation, which is in part influenced by AP-1 transcription factors. Thus, CAF-1 and SPT6 histone chaperones maintain cell fate through distinct mechanisms, highlighting how different chromatin assembly pathways can be leveraged to alter cell fate.

59

Chenghao Zhu

Abstract: Gene expression is a multi-step transformation of biological information from its storage form (DNA) into functional forms (protein and some RNAs). Regulatory activities at each step of this transformation multiply a single gene into a myriad of proteoforms. Proteogenomics is the study of how genomic and transcriptomic variation creates this proteomic diversity, and is limited by the challenges of modeling the complexities of gene expression. We therefore created moPepGen, a graph-based algorithm that comprehensively generates non-canonical peptides in linear time. moPepGen works with multiple technologies, in multiple species and on all types of genetic and transcriptomic data. In human cancer proteomes, it enumerates previously unobservable noncanonical peptides arising from germline and somatic genomic variants, noncoding open reading frames, RNA fusions and RNA circularization. We applied moPepGen to 376 cell lines, five prostate and eight kidney tumors, generating tumor-specific non-canonical peptides based on matched genomic and transcriptomic profiles. Using a tiered database search approach, we identified non-canonical peptides derived from disease-associated germline and somatic mutations, as well as tumor-specific transcriptomic variants, including fusions and circular RNAs. Additionally, we detected putative neoantigen peptides from common mutations such as KRAS and TP53. By enabling efficient identification and quantification of previously hidden proteins in both existing and new proteomic datasets, moPepGen facilitates all proteogenomics applications. It is available at: https://github.com/uclahs-cds/package-moPepGen.