Biology subjects

Lee, C. S.

Publications and source records attributed to Lee, C. S..

At least 19 recordsLinked to original sources

bioRxiv2026

Neural circuits regulating social dominance implement a strategy predicted by evolutionary game theory

Social conflict is a fundamental challenge for all animals and determines access to critical resources like mates and food. Evolutionary game theory predicts that natural selection should yield competitive strategies that balance the benefits and costs of social conflict. However, whether such strategies are embedded within the neural circuits that regulate aggression remains unclear. Here, we identify a neural circuit regulating the decision to flee during fighting in male Drosophila and show that the onset of defeat is governed by a probabilistic strategy predicted by evolutionary game theory. This mechanism arises from the inhibition of Tk-GAL4FruM neurons that promote aggressive arousal in males. Inhibition is mediated by a mushroom body circuit involving PPL1 dopaminergic neurons and V2 mushroom body output neurons, both classically associated with aversive learning. Silencing this circuit disrupts the onset of defeat, while activating it induces rapid defeat. Conversely, activation of reward-encoding PAM dopaminergic neurons promotes winning, revealing a dual role for dopamine in shaping contest dynamics. Finally, we find that internal state variables such as hunger and motivation shift the defeat onset probability distribution, consistent with game theory predictions of how payoff modulates fighting persistence. Together, our results provide direct evidence that evolutionary strategies based on payoff, long described by game theory, are implemented as circuit-level computations that regulate aggression.

neuroscience↗

bioRxiv2025

Dataset Documentation for Responsible AI: Analysis of Suitability and Usage for Health Datasets

Artificial Intelligence (AI) is rapidly transforming healthcare, but also raising concerns about algorithmic biases that mostly stem from the training data. It is widely supported that transparent dataset documentation is key to enabling responsible AI development. Several standardized dataset documentation approaches have been established, such as Datasheet, Dataset Nutrition Label, Accountability Documentation, Healthsheet, and Data Card. However, their suitability and usage for health datasets remain unclear. In this work, we compared all five approaches and evaluated their alignment with the STANDING Together Recommendations for Documentation of Health Datasets. We also investigated their real-world usage and gathered insights from generators and consumers of health datasets. Our findings reveal that none of these documentation approaches are used widely or fully suited for health datasets. We recommend developing a standard documentation approach for health datasets along with clear guidelines and automation tools to support adoption.

bioinformatics↗

bioRxiv2025

Reinforcement Learning for Antibody Sequence Infilling

We introduce a flexible framework for antibody sequence design that combines an infilling language model with reinforcement learning to optimize functional properties. Our approach leverages a pretrained infilling language model to generate specific antibody regions within full sequences, guided by reinforcement learning to improve desired biophysical characteristics. We implement a range of online learning strategies, exploring both vanilla REINFORCE and Proximal Policy Optimization with Kullback-Leibler (KL) regularization, and demonstrate that KL regularization is essential for maintaining a balance between score optimization and sequence plausibility. We also adapt Direct Reward Optimization to the protein domain by adding a value head to the infilling model, allowing it to learn directly from static (prompt, response, feedback) datasets using a mean-squared error objective. This formulation is particularly useful when only single-trajectory data is available, which is commonly the case for historically collected experimental assays. We evaluate both the online and offline methods across multiple antibody design tasks--including binding affinity, immunogenicity, and expression--and show that our framework improves alignment with measured biophysical properties while outperforming likelihood-only baselines. This integrated online/offline approach enables functionally driven antibody design and provides a scalable toolkit for therapeutic sequence engineering. Code and data are available at https://github.com/LLNL/protein_tune_rl.

bioinformatics↗

bioRxiv2025

Epithelial tumor cells utilize mast cell-derived histamine to regulate perineural invasion

Cancer dissemination by perineural invasion (PNI) is associated with poor outcomes in epithelial malignancies, yet the cellular interactions critical to this process are not fully defined. Single-cell transcriptomic analysis of cutaneous squamous cell carcinoma (cSCC) with PNI highlighted KITLG-KIT signaling activation between keratinocytes and mast cells. Additionally, we observed increased expression of HRH1 on tumor cells, which encodes the H1 histamine receptor. We confirmed these changes in other cancers, including PNI-positive hepatobiliary and oral mucosal tumors. Histamine enhanced tumor invasion into nerve in organoids that recapitulate key features of PNI, an effect rescued by H1-antihistamines. Mechanistically, broad upregulation of matrix metalloproteinases (MMPs) through P38 activation underscored histamine-mediated PNI by enabling collagen degradation in the nerve sheath. Examination of multiple cancer types with PNI showed co-expression of HRH1 and MMP2 on tumor cells as well as type IV collagen loss in involved nerves. Retrospective analysis of >15,000 patient health records in two independent databases demonstrated that H1-antihistamine use improved overall survival and immunotherapy response in cancers that spread by PNI. Together, these data reveal a critical role for mast cells in PNI and present a new therapeutic strategy for targeting this process in epithelial cancers.

cancer biology↗

bioRxiv2025

JNK pathway suppression drives resistance to combination endocrine therapy and CDK4/6 inhibition in ER+ breast cancer

PurposeEndocrine therapy in combination with CDK4/6 inhibition doubles the progression-free survival of patients with advanced ER+ breast cancer, but resistance is inevitable, leaving patients with limited treatment options. Experimental DesignWe performed unbiased genome-wide CRISPR/Cas9 knockout screens using ER+ breast cancer cells to identify novel drivers of resistance to combination endocrine therapy (tamoxifen) and CDK4/6 inhibitor (palbociclib) treatment. Screen hits were validated by CRISPR/Cas9 knockout models, mechanistic analyses and evaluation of patient samples. ResultsOur screens identified the inactivation of JNK signalling, including loss of the kinase MAP2K7, as a key driver of combination resistance. We developed multiple CRISPR/Cas9 knockout ER+ breast cancer cell lines (MCF-7 and T-47D) to investigate the effects of MAP2K7, MAPK8 and MAPK9 loss. MAP2K7 knockout increased metastatic burden in vivo and led to impaired JNK-mediated stress responses, as well as promoting cell survival and reducing senescence entry following endocrine therapy and CDK4/6 inhibitor treatment. Mechanistically, this occurred via loss of the AP-1 transcription factor c-JUN, leading to an attenuated response to combination endocrine therapy plus CDK4/6 inhibition. Furthermore, we analysed ER+ advanced breast cancer patient cohorts and found that inactivation of the JNK pathway was associated with increased metastatic burden, and low pJNKT183/Y185 activity correlated with a poorer response to systemic endocrine and CDK4/6 inhibitor therapies. ConclusionsOverall, we demonstrate that suppression of JNK signalling enables persistent growth during combined endocrine therapy and CDK4/6 inhibition. Furthermore, our data provide a pre-clinical rationale to screen patients tumours for JNK signalling deficiency prior to receiving combined endocrine therapy and CDK4/6 inhibition. STATEMENT OF TRANSLATIONAL RELEVANCEResistance to CDK4/6 inhibitors in the context of endocrine therapy resistance presents an urgent clinical challenge for the management of estrogen receptor positive (ER+) breast cancer. However, the mechanisms driving resistance to this therapeutic combination remain poorly understood. Here, we have identified inactivation of JNK signalling, specifically loss of the JNK kinase MAP2K7, as a major determinant of resistance to combined endocrine therapy and CDK4/6 inhibition in ER+ breast cancer. We uncovered that MAP2K7 loss augments tumour survival in vitro and in vivo. This occurs via disruption of activator protein-1 (AP-1) transcription factors, and thus prevents the induction of therapy-induced senescence and JNK-induced stress response. These findings reveal a critical tumour suppressor role for the JNK pathway in ER+ breast cancer, highlighting the importance of identifying patients with deficient JNK signalling and cautioning against the development of JNK inhibitors for this setting.

cancer biology↗

bioRxiv2024

Inhibition of serine and arginine-rich splicing factor 3 induces epidermal differentiation and decreases cutaneous squamous cell carcinoma risk

Serine/Arginine-rich splicing factor 3 (SRSF3) is one of 12 SRSFs that regulate gene expression via alternative splicing. SRSF3 is upregulated in cutaneous squamous cell carcinoma (cSCC) and several squamous cancer cell lines in relation to normal keratinocytes. We suppressed SRSF3 with a specific inhibitor, SFI003, and observed an increase in epidermal differentiation. Our data suggests that in cSCC, SRSF3 overexpression suppresses cellular differentiation to enable cancer progression. In a clinical setting, patients taking known SRSF3 inhibitors digoxin and amiodarone exhibited higher cSCC-free survival compared to a propensity score-matched cohort treated with beta blockers. Thus, SRSF3 upregulation may be a novel therapeutic target in cSCC that can improve patient prognoses.

cancer biology↗

bioRxiv2024

Neural and verbal responses to attachment-schema narratives differ based on past and current caregiving experiences

We investigated how past and current caregiving experiences impacted emotional event processing by examining inter-subject functional correlation in 7- to 15-year-olds during narrative movies depicting separation and reunion. Early adversity impacted amygdala interactions with the hippocampus, but medial prefrontal-amygdala connectivity and the content of later recall was shaped by attachment security to current caregivers. This suggests that current attachment can have a redemptive effect on attachment schemas formed in early childhood.

neuroscience↗

bioRxiv2024

Extrachromosomal DNA in cutaneous squamous cell carcinoma is associated with increased nodal disease

Extrachromosomal DNA (ecDNA) is a mechanism of oncogene amplification that occurs in many cancers and has been linked to worse patient outcomes. While ecDNA has previously been identified in cutaneous melanoma, it has yet to be described in cutaneous squamous cell carcinoma (cSCC). Using whole genome sequencing and transcriptomic data of primary and metastatic cSCC tumors, we establish the presence of ecDNA in cSCC and confirm elevated expression of ecDNA-associated genes. We also find that the presence of ecDNA correlates with a higher lymph node ratio, a measure of disease severity. Together, these findings are the first to report ecDNA in cSCC and suggest that ecDNA gene amplifications promote an aggressive cSCC phenotype.

cancer biology↗

bioRxiv2023

Gene expression landscape of cutaneous squamous cell carcinoma progression

BackgroundCutaneous squamous cell carcinomas (SCC) are the second most common human cancer and have been characterized by RNA sequencing (RNA-Seq); however, the transferability of findings from individual studies may be limited by small sample sizes and diverse analysis protocols. ObjectivesTo define the transcriptome landscape at different stages in the progression of normal skin to SCC through a meta-analysis of publicly available RNA-Seq samples MethodsWhole-transcriptome data from 73 normal skin samples, 46 actinic keratoses (AK), 16 in situ SCC, 13 keratoacanthomas (KA), and 147 SCC (including 30 SCC from immunocompromised patients and 8 SCC from individuals with recessive dystrophic epidermolysis bullosa [RDEB]) was uniformly processed to harmonize gene expression. Differential expression, fusion detection, and cell-type deconvolution analyses were performed. ResultsIndividual RNA-Seq studies of SCC demonstrated study-specific clustering and varied widely in their differential gene expression detection. Following batch correction, we defined a consensus set of differentially expressed genes (DEGs), including those altered in the preinvasive stages of SCC development, and used single-cell RNA-Seq data to demonstrate that DEGs are often, but not always, expressed by tumor-specific keratinocytes (TSKs). Analysis of the cellular composition of SCC, KA, and RDEB-SCC identified an increase in differentiated keratinocytes in KA, while RDEB-SCC contained the most TSKs. Compared to SCC arising in immunocompetent patients, SCC from immunosuppressed individuals demonstrated fewer memory B cells and CD8 T cells. A comprehensive and unbiased search for fusion transcripts in SCC and intermediate disease stages identified few candidates that recur in >1% of all specimens, suggesting most SCC are not driven by oncogenic gene fusions. Finally, using GTEx data, we distilled a novel 300-gene signature of chronic sun exposure that affirms greater cumulative ultraviolet (UV) exposure in later stages of SCC development. ConclusionsOur results define the gene expression landscape of SCC progression, characterize cell subpopulation heterogeneity in SCC subtypes that contribute to their distinct clinical phenotypes, demonstrate that gene fusions are not a common cause of SCC, and identify UV-responsive genes associated with SCC development. What is already known about this topic?O_LICutaneous squamous cell carcinoma (SCC) is one of the most common cancers worldwide. C_LIO_LISeveral studies have examined gene expression changes in SCC using RNA sequencing (RNA-Seq) but comparison of their results is difficult due to inter-study variation and diverse bioinformatic pipelines and protocols. C_LIO_LIA few gene fusions have been described in SCC, but a comprehensive characterization of fusion transcripts in patient samples has not been performed. C_LI What does this study add?O_LIWe re-analyzed RNA-Seq data from 11 studies of SCC and its preinvasive stages to create a list of consensus differentially expressed genes and identify those that are UV-responsive. C_LIO_LIClinically aggressive SCC displayed more tumor-specific keratinocytes, while keratoacanthomas contained more differentiated keratinocytes. SCC in immunocompetent persons had more memory B cells and CD8 T cells than those arising in immunosuppressed individuals. C_LIO_LIPreviously reported gene fusions were not detected and most fusion candidates did not demonstrate pathogenic features. C_LI What is the translational message?O_LIOur analysis harmonizes differing results from previous studies to provide a robust list of genes implicated in SCC development. C_LIO_LIOur findings suggest gene fusions are not a common driver event in SCC. C_LI

cancer biology↗

bioRxiv2023

Specific and non-linear effects of glaucoma on optic radiation tissue properties

Changes in sensory input with aging and disease affect brain tissue properties. To establish the link between glaucoma, the most prevalent cause of irreversible blindness, and changes in major brain connections, we characterized white matter tissue properties in diffusion MRI measurements in a large sample of subjects with glaucoma (N=905; age 49-80) and healthy controls (N=5,292; age 45-80) from the UK Biobank. Confounds due to group differences were mitigated by matching a sub-sample of controls to glaucoma subjects. A convolutional neural network (CNN) accurately classified whether a subject has glaucoma using information from the primary visual connection to cortex (the optic radiations, OR), but not from non-visual brain connections. On the other hand, regularized linear regression could not classify glaucoma, and the CNN did not generalize to classification of age-group or of age-related macular degeneration. This suggests a unique non-linear signature of glaucoma in OR tissue properties.

neuroscience↗

bioRxiv2022

Optic radiations representing different eccentricities age differently

The white matter pathways that carry information from the foveal, macular, and peripheral visual fields have distinct biological properties. The optic radiations (OR) carry foveal and peripheral information from the thalamus to the primary visual cortex (V1) through adjacent but separate pathways in the white matter. Here, we perform white matter tractometry using pyAFQ 1 on a large sample of diffusion MRI (dMRI) data from the UK Biobank dataset (UKBB; N=6021; age 45-81). We use pyAFQ to characterize white matter tissue properties in parts of the OR that transmit information about the foveal, macular, and peripheral visual fields, and to characterize the changes in these tissue properties with age. We find that (1) independent of age there is higher fractional anisotropy, lower mean diffusivity, and higher mean kurtosis in the foveal and macular OR than in peripheral OR, consistent with denser, more organized nerve fiber populations in foveal/parafoveal pathways, and (2) age is associated with increased diffusivity and decreased anisotropy and kurtosis, consistent with decreased density and tissue organization with adulthood aging. However, anisotropy in foveal OR decreases faster with age than in peripheral OR, while diffusivity increases faster in peripheral OR, suggesting foveal/peri-foveal OR and peripheral OR differ in how they age.

neuroscience↗

bioRxiv2022

RET activation controlled by MAB21L4-CacyBP interaction drives squamous cell carcinoma

Epithelial squamous cell carcinomas (SCC) most commonly originate in the skin, where they display disruptions in the normally tightly regulated homeostatic balance between keratinocyte proliferation and terminal differentiation. We performed a transcriptome-wide screen for genes of unknown function that possess inverse expression patterns in differentiating keratinocytes compared to cutaneous SCC (cSCC) and identified MAB21L4 (C2ORF54) as an enforcer of terminal differentiation that suppresses carcinogenesis. Loss of MAB21L4 in human cSCC organoids enabled malignant transformation through increased expression of the receptor tyrosine kinase rearranged during transfection (RET). In addition to transcriptional upregulation of RET, MAB21L4 deletion preempted recruitment of the CacyBP-Siah1 E3 ligase complex to RET and reduced its ubiquitylation. Both genetic disruption of RET or selective RET inhibition with BLU-667 (pralsetinib) suppressed tumorigenesis in SCC organoids and in vivo tumors while inducing concomitant differentiation. Our results suggest that targeting RET activation is a potential therapeutic strategy for treating SCC. Statement of SignificanceFew targeted therapies are available to individuals with cSCC who seek or require non-surgical management. Our study demonstrates that downregulation of RET is required for epithelial differentiation and opposes carcinogenesis in cSCC as well as SCC arising from other epithelial tissues.

cancer biology↗

bioRxiv2021

Network diffusion model predicts neurodegeneration in limb-onset amyotrophic lateral sclerosis

The trans-neural propagation of phosphorylated 43-kDa transactive response DNA-binding protein (pTDP-43) contributes to neurodegeneration in Amyotrophic Lateral Sclerosis (ALS). We investigated whether Network Diffusion Model (NDM), a biophysical model of spread of pathology via the brain connectome, could capture the severity and progression of neurodegeneration (atrophy) in ALS. We measured degeneration in limb-onset ALS patients (n=14 at baseline, 12 at 6-months, and 9 at 12 months) and controls (n=12 at baseline) using FreeSurfer analysis on the structural T1-weighted Magnetic Resonance Imaging (MRI) data. The NDM was simulated on the canonical structural connectome from the IIT Human Brain Atlas. To determine whether NDM could predict the atrophy pattern in ALS, the accumulation of pathology modelled by NDM was correlated against atrophy measured using MRI. The cross-sectional analyses revealed that the network diffusion seeded from the inferior frontal gyrus (pars triangularis and pars orbitalis) significantly predicts the atrophy pattern in ALS compared to controls. Whereas, atrophy over time with-in the ALS group was best predicted by seeding the network diffusion process from the inferior temporal gyrus at 6-month and caudal middle frontal gyrus at 12-month. Our findings suggest the involvement of extra-motor regions in seeding the spread of pathology in ALS. Importantly, NDM was able to recapitulate the dynamics of pathological progression in ALS. Understanding the spatial shifts in the seeds of degeneration over time can potentially inform further research in the design of disease modifying therapeutic interventions in ALS.

neuroscience↗

bioRxiv2020

Anticipation of temporally structured events in the brain

1Learning about temporal structure is adaptive because it enables the generation of expectations. We examined how the brain uses experience in structured environments to anticipate upcoming events. During fMRI, individuals watched a 90-second movie clip six times. Using a Hidden Markov Model applied to searchlights across the whole brain, we identified temporal shifts between activity patterns evoked by the first vs. repeated viewings of the movie clip. In many regions throughout the cortex, neural activity patterns for repeated viewings shifted to precede those of initial viewing by up to 15 seconds. This anticipation varied hierarchically in a posterior (less anticipation) to anterior (more anticipation) fashion. We also identified specific regions in which the timing of the brains event boundaries were related to those of human-labeled event boundaries, with the timing of this relationship shifting on repeated viewings. With repeated viewing, the brains event boundaries came to precede human-annotated boundaries by 1-4 seconds on average. Together, these results demonstrate a hierarchy of anticipatory signals in the human brain and link them to subjective experiences of events.

neuroscience↗

bioRxiv2020

Integrated hiPSC-based liver and heart microphysiological systems predict unsafe drug-drug interaction

Microphysiological systems (MPSs) mimicking human organ function in vitro are an emerging alternative to conventional cell culture and animal models for drug development. Human induced pluripotent stem cells (hiPSCs) have the potential to capture the diversity of human genetics and provide an unlimited supply of cells. Combining hiPSCs with microfluidics technology in MPSs offers new perspectives for drug development. Here, the integration of a newly developed liver MPS with a cardiac MPS--both built with the same hiPSC line--to study drug-drug interaction (DDI) is reported. As a prominent example of clinically relevant DDI, the interaction of the arrhythmogenic gastroprokinetic cisapride with the fungicide ketoconazole was investigated. As seen in patients, metabolic conversion of cisapride to non-arrhythmogenic norcisapride in the liver MPS by the cytochrome P450 enzyme CYP3A4 was inhibited by ketoconazole, leading to arrhythmia in the cardiac MPS. These results establish functional integration of isogenic hiPSC-based liver and cardiac MPSs, which allows screening for DDI, and thus drug efficacy and toxicity, in the same genetic background.

bioengineering↗

bioRxiv2019

High-phytate/low-calcium diet is a risk factor for crystal nephropathies, renal phosphate wasting, and bone loss

Phosphate overload contributes to mineral bone disorders associated with crystal nephropathies. Phytate, the major form of phosphorus in plant seeds, is known as an indigestible and negligible in humans. However, the mechanism and adverse effects of high-phytate intake on Ca2+ and phosphate absorption and homeostasis are unknown. Here we show that excessive intake of phytate with a low-Ca2+ diet fed to rats contributed to the development of crystal nephropathies, renal phosphate wasting, and bone loss through tubular dysfunction secondary to dysregulation of intestinal calcium and phosphate absorption. Moreover, Ca2+ supplementation alleviated the detrimental effects of excess dietary phytate on bone and kidney through excretion of undigested Ca2+-phytate, which prevented a vicious cycle of intestinal phosphate overload and renal phosphate wasting while improving intestinal Ca2+ bioavailability. Thus, we demonstrate that phytate is digestible without a high-Ca2+ diet and a risk factor for phosphate overloading and developing crystal nephropathies and bone disease.

physiology↗

bioRxiv2018

Forecasting Future Humphrey Visual Fields Using Deep Learning

PurposeTo determine if deep learning networks could be trained to forecast a future 24-2 Humphrey Visual Field (HVF).\n\nDesignRetrospective database study.\n\nParticipantsAll patients who obtained a HVF 24-2 at the University of Washington.\n\nMethodsAll datapoints from consecutive 24-2 HVFs from 1998 to 2018 were extracted from a University of Washington database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction.\n\nMain outcome measuresMean absolute error (MAE) and difference in Mean Deviation (MD) between predicted and actual future HVF.\n\nResultsMore than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24-2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall MAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained models were able to successfully predict progressive field loss in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF (p < 2.2 = 10-16) and an average difference of 0.41 dB.\n\nConclusionsUsing unfiltered real-world datasets, deep learning networks show an impressive ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

bioinformatics↗

bioRxiv2018

Generating perfusion maps from structural optical coherence tomography with artificial intelligence

Despite advances in artificial intelligence (AI), its application in medical imaging has been burdened and limited by expert-generated labels. We used images from optical coherence tomography angiography (OCTA), a relatively new imaging modality that measures retinal blood flow, to train an AI algorithm to generate flow maps from standard optical coherence tomography (OCT) images, exceeding the ability and bypassing the need for expert labeling. Deep learning was able to infer flow from single structural OCT images with similar fidelity to OCTA and significantly better than expert clinicians (P < 0.00001). Our model allows generating flow maps from large volumes of previously collected OCT data in existing clinical trials and clinical practice. This finding demonstrates a novel application of AI to medical imaging, whereby subtle regularities between different modalities are used to image the same body part and AI is used to generate detailed inferences of tissue function from structure imaging.

neuroscience↗