Biology subjects

Choi, L.

Publications and source records attributed to Choi, L..

6 recordsLinked to original sources

bioRxiv2026

Representational geometry of punishment learning across basolateral amygdala neurons

Punishment learning allows animals to suppress actions that cause harm while maintaining actions that earn reward. Although the basolateral amygdala is necessary for punishment learning, how its population activity changes across learning, and how this activity reflects an action's consequences, remain unknown. We used longitudinal calcium imaging to track 1,147 basolateral amygdala neurons in male mice across instrumental training, punishment, choice test, and punishment extinction. Punishment learning transformed basolateral amygdala neuron activity associated with the punished action. This activity diverged from that associated with the unpunished action and became aligned with activity evoked by the punisher itself. This alignment was action and contingency specific. It preserved the direction of individual neurons' responses to the punisher: neurons activated by the punisher become more activated by the punished action alone, whereas neurons inhibited by the punisher become more inhibited by the punished action alone. We then asked whether reward and punishment were encoded separately or on a common value dimension across basolateral amygdala neurons. Punishment added a strong shock component to action-related activity without reducing its reward component, while reward and shock themselves occupied largely orthogonal population directions. Thus, during punishment learning, basolateral amygdala population activity can encode opposing consequences of the same action instead of simply changing what that action is worth.

neuroscience↗

bioRxiv2025

BDNF receptor balance gates AgRP-to-PVH fiber density and stimulation evoked feeding across caloric states

Agouti-related peptide (AgRP) neurons show opposite pairings of anatomy and output across caloric states. In mice, 24-h caloric restriction (CR) increases AgRP-to-paraventricular hypothalamus (PVH) fiber density without increasing stimulation-evoked feeding, whereas 5-day high-fat diet (HFD) reduces fiber density yet increases stimulation-evoked feeding, including during PVH terminal stimulation. PVH Bdnf rises acutely under both challenges and remains elevated with sustained HFD. Receptor expression shifts with state: in AgRP neurons, CR increases TrkB transcripts, whereas HFD increases p75NTR transcripts. AgRP-specific deletions define directional roles: TrkB loss reduces AgRP-to-PVH fiber density and yields larger stimulation-evoked feeding, whereas p75NTR loss increases fiber density and limits responses. Thus, a TrkB/p75NTR balance in AgRP neurons provides rheostat-like control over the relationship between circuit structure and stimulation-evoked feeding across caloric states, identifying TrkB and p75NTR as tractable molecular handles for probing state-dependent control of AgRP circuit output. HighlightsO_LICR increases AgRP[->]PVH fibers without boosting evoked feeding C_LIO_LIHFD retracts AgRP[->]PVH fibers and amplifies evoked feeding C_LIO_LIPVH Bdnf rises acutely in response to CR or HFD; elevation persists only with HFD C_LIO_LIAgRP TrkB{uparrow} with CR, p75NTR{uparrow} with HFD; AgRP specific knockouts of these receptors define floor and ceiling PVH axon innervation/evoked feeding C_LI

neuroscience↗

bioRxiv2025

Disturbed Flow Induces Reprogramming of Endothelial Cells to Immune-like and Foam Cells under Hypercholesterolemia during Atherogenesis

BackgroundAtherosclerosis occurs preferentially in the arteries exposed to disturbed flow (d-flow), while the stable flow (s-flow) regions are protected even under hypercholesterolemic conditions. We recently showed that d-flow alone initiates flow-induced reprogramming of endothelial cells (FIRE), including the novel concept of partial endothelial-to-immune-cell-like transition (partial EndIT), but was not validated using a genetic lineage-tracing model. Here, we tested and validated the two-hit hypothesis that d-flow is an initial instigator of partial FIRE but requires hypercholesterolemia to induce a full-blown FIRE and atherosclerotic plaque development. MethodsMice were treated with adeno-associated virus expressing proprotein convertase subtilisin/kexin type 9 and a Western diet to induce hypercholesterolemia and/or partial carotid ligation (PCL) surgery to expose the left common carotid artery (LCA) to d-flow. Single-cell RNA sequencing (scRNA-seq) analysis was performed using cells obtained from the intima and leftover LCAs and the control right common carotid arteries at 2 and 4 weeks post-PCL. Comprehensive immunohistochemical staining was performed on EC-specific confetti mice treated with PCL and hypercholesterolemic conditions at 4 weeks post-PCL to validate endothelial reprogramming. ResultsAtherosclerotic plaques developed by d-flow under hypercholesterolemia at 2 and 4 weeks post-PCL, but not by d-flow or hypercholesterolemia alone, as expected. The scRNA-seq results of 98,553 single cells from 95 mice revealed 25 cell clusters; 5 EC, 3 vascular smooth muscle cell (SMC), 5 macrophage (M{Phi}), and additional fibroblast, T cell, natural killer cell, dendritic cell, neutrophil, and B cell clusters. Our scRNA-seq analyses showed that d-flow under hypercholesterolemia transitioned healthy ECs to full immune-like (EndIT) and, more surprisingly, foam cells (EndFT), in addition to inflammatory and mesenchymal cells (EndMT). Further, EC-derived foam cells shared remarkably similar transcriptomic profiles with foam cells derived from SMCs and M{Phi}s. Comprehensive lineage-tracing studies using immunohistochemical staining of canonical protein and lipid markers in the EC-specific confetti mice clearly demonstrated direct evidence supporting the novel FIRE hypothesis, including EndIT and EndFT, when d-flow was combined with hypercholesterolemia. Further, reanalysis of the publicly available human carotid plaque scRNA-seq and Perturb-seq datasets supported the FIRE hypothesis and a potential mechanistic link between the genes and FIRE. ConclusionWe provide evidence supporting the two-hit hypothesis: ECs in d-flow regions, such as the branching points, are partially reprogrammed, while hypercholesterolemia alone has minimal endothelial reprogramming effects. Under hypercholesterolemia, d-flow fully reprograms arterial ECs, including the novel EndIT and EndFT, in addition to inflammation and EndMT, during atherogenesis. This single-cell atlas provides a crucial roadmap for developing novel mechanistic understanding and therapeutics targeting flow-sensitive genes, proteins, and pathways of atherosclerosis.

cell biology↗

bioRxiv2021

Statistical and machine learning models for classification of human wear and delivery days in accelerometry data

PurposeAccelerometers are increasingly utilized in healthcare research to assess human activity. Accelerometry data are often collected by mailing accelerometers to participants, who wear the accelerometers to collect data on their activity. The devices are then mailed back to the laboratory for analysis. We develop models to classify days in accelerometry data as activity from actual human wear or the delivery process. These models can be used to automate the cleaning of accelerometry datasets that are adulterated with activity from delivery. MethodsFor the classification of delivery days in accelerometry data, we developed statistical and machine learning models in a supervised learning context using a large human activity and delivery labeled accelerometry dataset. We extracted several features, which were included to develop random forest, logistic regression, mixed effects regression, and multilayer perceptron models, while convolutional neural network, recurrent neural network, and hybrid convolutional recurrent neural network models were developed without feature extraction. Model performances were assessed using Monte Carlo cross-validation. ResultsWe found that a hybrid convolutional recurrent neural network performed best in the classification task with an F1 score of 0.960 but simpler models such as logistic regression and random forest also had excellent performance with F1 scores of 0.951 and 0.957, respectively. ConclusionThe models developed in this study can be used to classify days in accelerometry data as either human or delivery activity. An analyst can weigh the larger computational cost and greater performance of the convolutional recurrent neural network against the faster but slightly less powerful random forest or logistic regression. The best performing models for classification of delivery data are publicly available on the open source R package, PhysicalActivity.

bioinformatics↗

bioRxiv2019

A post-processing algorithm for building longitudinal medication dose data from extracted medication information using natural language processing from electronic health records

ObjectiveWe developed a post-processing algorithm to convert raw natural language processing (NLP) output from electronic health records (EHRs) into a usable format for analysis. This algorithm was specifically developed for creating datasets for use in medication-based studies. Materials and MethodsThe algorithm was developed using output from two NLP systems, MedXN and medExtractR. We extracted medication information from deidentified clinical notes from Vanderbilts EHR system for two medications, tacrolimus and lamotrigine. The algorithm consists of two parts. Part I parses the raw NLP output and connects entities together. Part II removes redundancies and calculates dose intake and daily dose. We evaluated each part by comparing to human-determined gold standards that were generated using approximately 300 records from 10 subjects for each medication and each NLP system. ResultsThe algorithm performed well. For MedXN, the F-measures were at or above 0.99 for Part I and at or above 0.97 for Part II. For medExtractR, the F-measures for Part I were 1.00 and for Part II they were at or above 0.98. DiscussionOur post-processing algorithm was developed separately from an NLP system, making it easier to modify and generalize to other systems. It performed well to convert NLP output to analyzable data, but it cannot perform well in certain cases, such as when incorrect information is extracted by the NLP system. ConclusionOur post-processing algorithm provides a way to convert raw NLP output to a form that is useful for medication-based studies, leading to more opportunities to use EHR data for diverse studies.

bioinformatics↗

bioRxiv2017

How Well Do You Know Your Mutation? Complex Effects Of Genetic Background On Expressivity, Complementation, And Ordering Of Allelic Effects

For a given gene, different mutations influence organismal phenotypes to varying degrees. However, the expressivity of these variants not only depends on the DNA lesion associated with the mutation, but also on factors including the genetic background and rearing environment. The degree to which these factors influence related alleles, genes, or pathways similarly, and whether similar developmental mechanisms underlie variation in the expressivity of a single allele across conditions and variation across alleles is poorly understood. Besides their fundamental biological significance, these questions have important implications for the interpretation of functional genetic analyses, for example, if these factors alter the ordering of allelic series or patterns of complementation. We examined the impact of genetic background and rearing environment for a series of mutations spanning the range of phenotypic effects for both the scalloped and vestigial genes, which influence wing development in Drosophila melanogaster. Genetic background and rearing environment influenced the phenotypic outcome of mutations, including intra-genic interactions, particularly for mutations of moderate expressivity. We examined whether cellular correlates (such as cell proliferation during development) of these phenotypic effects matched the observed phenotypic outcome. While cell proliferation decreased with mutations of increasingly severe effects, surprisingly it did not co-vary strongly with the degree of background dependence. We discuss these findings and propose a phenomenological model to aid in understanding the biology of genes, and how this influences our interpretation of allelic effects in genetic analysis.

genetics↗