Biology subjects

Shin, H.

Publications and source records attributed to Shin, H..

At least 19 recordsLinked to original sources

bioRxiv2026

Towards Robust Automated Mouse Scratching Behavior Detection for Itch Research

Automated detection of mouse scratching behavior could reduce manual annotation and is essential for preclinical itch research. However, robust performance across laboratories remains a challenge for current methods. We present Scratch-AID+, a framework dedicated to automated mouse scratching detection, including a diverse and progressive set of approaches: classic ResNet-based detectors, task-adaptive fine-tuning of video foundation models, and in-context learning with advanced multimodal large language models. To simulate real-world use cases, we benchmark on a public dataset and utilize additional cross-laboratory video recordings for independent testing. We find that classic baselines suffered severe performance degradation on external data under the realistic setting with sparse scratching events. In contrast, fine-tuned video foundation models successfully bridged this generalizability gap, achieving significantly better detection performance. Furthermore, frozen multimodal large language models demonstrated promising out-of-the-box ability using only few-shot demonstrations. We also examined the effects of different video preprocessing strategies. Our findings highlight the improved performance and generalizability of the adapted foundation models for automated scratching behavior analysis and shed light on the real-world deployment of such detectors. Our code is publicly available at https://anonymous.4open.science/r/scratch aid-plus/.

bioinformatics↗

bioRxiv2026

De novo design of protease-activatable cytokine prodrugs

Cytokines can elicit potent antitumor immunity, but their clinical use is often limited by systemic immune activation and dose-limiting toxicity. Protease-responsive masking offers a way to improve selectivity by linking cytokine activity to disease-associated protease cleavage, but current approaches have faced several challenges. Here we describe a de novo protein design strategy for generating protease-activatable cytokine prodrugs that overcomes these challenges. We use this strategy to design and characterize cytokine prodrugs and well-behaved AND-gated split systems that require both target-dependent colocalization and proteolytic unmasking to generate active cytokines. In challenging syngeneic tumor models, our IL-21 prodrugs reduced treatment-associated toxicity while maintaining antitumor efficacy.

bioengineering↗

bioRxiv2026

Thal-Kak: unifying biomolecular structure predictors reveals a sampling-selection gap

Complementary all-atom structure predictors sample different solutions, but how to allocate a fixed sampling budget across them and select the best output remains unclear. Thal-Kak unifies five released predictors under shared upstream inputs and a common schema. Across FoldBench and CASP16, model mixing improves oracle sampling over single-model runs, but selection remains a bottleneck because confidence scores do not transfer across models and existing quality-assessment methods cannot resolve this gap.

bioinformatics↗

bioRxiv2026

Multicellular Programs Associated with Right Ventricular Adaptation in Pulmonary Arterial Hypertension

BackgroundRight ventricular (RV) adaptation determines outcomes in pulmonary arterial hypertension (PAH), yet multicellular molecular programs associated with adaptive versus maladaptive RV remodeling in living humans remain incompletely defined. MethodsWe collected 32 human RV tissue biopsies from patients with idiopathic PAH, systemic sclerosis-associated PAH (SSc-PAH), systemic sclerosis without pulmonary hypertension, with 24 nonfailing donor RVs serving as controls. We performed single-nucleus RNA sequencing and integrated cell-type specific transcriptional programs with contemporaneously obtained multi-beat pressure-volume loop measurements of RV contractility (Ees, end-systolic elastance) and RV-pulmonary arterial coupling (the ratio of Ees to Ea, the effective arterial load). SSc modification of PAH-associated biology was assessed using interaction terms. Bulk RV proteomic pathway enrichment was performed to assess an orthogonal molecular layer, and exploratory cell-cell communication analyses alongside independent spatial transcriptomic analyses were performed to contextualize key findings. ResultsPAH was associated with broad depletion of biosynthetic, trafficking, and mitochondrial programs across cell types. SSc modified the magnitude of many PAH-associated transcriptional programs while largely preserving pathway directionality. Significant multicellular pathway enrichments were associated with RV-pulmonary arterial coupling. Joint analysis of Ees, Ea, and Ees/Ea identified biologic programs associated with different RV responses to varying loading conditions. Preserved coupling was characterized by enriched extracellular matrix, laminin-integrin, receptor tyrosine kinase, mitochondrial, and translational programs involving fibroblast, endothelial, endocardial, and cardiomyocyte compartments. Cell- cell communication analyses predicted coordinated stromal-vascular signaling networks involving laminin-integrin and endothelial-to-mural signaling in preserved coupling. Proteomic and spatial analyses supported recurrent multicellular themes. ConclusionsRV adaptation in PAH is associated with distinct, coordinated multicellular programs that vary with load and contractile response. RV-PA coupling in PAH is associated with multicellular remodeling that extends beyond cardiomyocytes and reflects organized vascular-stromal support architecture. These findings identify extracellular matrix, laminin- integrin signaling, mitochondrial, and translational programs as associated with adaptive RV remodeling in PAH. Clinical PerspectiveO_ST_ABSWhat is new?C_ST_ABSO_LICell type-resolved molecular profiling of living human RV tissue identifies PAH-associated depletion of biosynthetic, trafficking, mitochondrial, and repair-associated programs across multiple cardiac cell types. C_LIO_LIIntegration with contemporaneously obtained pressure-volume loop physiology demonstrates that preserved RV-pulmonary arterial coupling under lower load was associated predominantly with cardiomyocyte mitochondrial and metabolic competency, whereas preserved coupling under higher load was associated with extracellular matrix remodeling and vascular-stromal signaling. C_LIO_LIProteomic, cell-cell communication, and spatial analyses provided orthogonal support for coordinated extracellular matrix and vascular-stromal programs associated with preserved RV-pulmonary arterial coupling. C_LI What are the clinical implications?O_LIThese findings shift the biology of RV adaptation from a predominantly cardiomyocyte- centered model toward a multicellular tissue model in which metabolic, matrix, and vascular support programs vary according to loading conditions and contractile states. C_LIO_LIExtracellular matrix-integrin signaling, endothelial-mural communication, and mitochondrial competency represent candidate pathways for mechanistic investigation toward RV-directed therapies in PAH. C_LI

molecular biology↗

bioRxiv2026

Domain-specific proteome remodeling defines mouse myelin aging

Myelin, once regarded as a static insulating structure, is now recognized as a dynamic component of the nervous system, capable of remodeling in response to experience. Its breakdown is linked to cognitive decline and neurodegenerative diseases, often preceding neuronal dysfunction. While the myelin proteome has been studied, the age- and sex-related changes it undergoes remain poorly understood. In this study, we purified myelin proteins from young, middle-aged, and aged male and female mice. Using mass spectrometry-based proteomics, we identified 4,095 unique proteins in males and 3,931 in females, with roughly 30% showing significant age-related changes in both sexes. We find an age-related increase in compact myelin structural proteins, such as MBP, MOBP, and CLDN11, and a selective vulnerability in non-compact myelin cytoskeletal proteins, such as SEPTIN2, SEPTIN8, and OPALIN. Notably, disease-associated proteins previously characterized in single-cell transcriptomics appear at the protein level in aged myelin. To firmly distinguish between proteins derived from oligodendrocytes and other cell types in vivo we labeled nascent oligodendrocyte proteins with bio-orthogonal non-canonical amino acid tagging (BONCAT). We discovered a dramatic age-related increase in lysosomal and vesicle-associated proteins, while translation and synaptic proteins decrease in myelin with age in both sexes. This dataset highlights molecular mechanisms underlying the loss of myelin integrity and function with age and provides a novel tool for studying oligodendrocyte-derived nascent proteins in vivo.

neuroscience↗

bioRxiv2026

Continuous negative autoregulation fine-tunes dosage-sensitive transcription factor expression to maintain post-mitotic neuron identity

How post-mitotic neurons maintain precise transcription factor (TF) levels throughout life remains a fundamental open question. Here, we challenge the prevailing model of positive autoregulation by demonstrating that UNC-3 (Collier/EBF1-4), a dosage-sensitive TF continuously required for cholinergic motor neuron identity in C. elegans, negatively regulates its own expression. Using genetics, biochemistry, and inducible protein depletion, we show this self-repression occurs directly at the transcriptional level and persists beyond development. CRISPR/Cas9 disruption of negative autoregulation causes motor neuron identity and locomotion defects, establishing its functional necessity. Mechanistically, the UNC-3 DNA-binding domain is required and sufficient for self-repression, with an AlphaFold2 screen implicating chromatin factors as interaction partners. Critically, UNC-3 self-repression is continuously counterbalanced by positive input from the HOX cofactor CEH-20/PBX, revealing a dynamic "balancing act" between opposing regulatory inputs that stabilize TF dosage over time. Mutations in the unc-3 ortholog EBF3 cause a neurodevelopmental syndrome, and disease-associated variants disrupt UNC-3 self-repression, revealing a key molecular mechanism underlying the disorder. We propose that negative autoregulation continuously counteracted by positive input represents a broadly applicable principle for maintaining dosage-sensitive TF expression to secure post-mitotic cell identity.

developmental biology↗

bioRxiv2026

CDS-BART: A BART-Based Foundation Model for mRNA Sequence Analysis

Summary: Recent advancements in artificial intelligence (AI) have led to the development of foundation models that interpret mRNA as a language. Notable examples include CodonBERT, hydraRNA, EVO2, and Helix-mRNA. These models demonstrate significant potential as powerful tools for mRNA research. However, to best of our knowledge, there is currently no publicly available AI model that is both easy to use and capable of analyzing mRNA sequences up to about 4kb, a length scale typical of many therapeutic mRNAs, including those encapsulated within lipid nanoparticls (LNPs). Thus, we propose CDS-BART, a user-friendly, open-source tool that integrates SentencePiece sub-word tokenization with the denoising sequence-to-sequence training of Bidirectional and Auto-Regressive Transformers (BART). CDS-BART was pre-trained on mRNA data from nine taxonomic groups provided by the NCBI RefSeq database. This comprehensive pre-training, coupled with BARTs denoising capability, ensures effective learning of codon usage, mRNA structure, evolution, and regulation. Thus, CDS-BART can ultimately deliver robust performance across a wide range of mRNA prediction tasks. Availability and ImplementationCDS-BART is released under the MIT License. Latest code is available via Github at https://github.com/mogam-ai/CDS-BART.

bioinformatics↗

bioRxiv2026

Pancreatic cancer-associated organ dysfunction promotes muscle autophagy and contributes to peripheral tissue wasting

Normal pancreas function supports both digestion and the hormonal regulation of whole-body metabolism. We find pancreatic ductal adenocarcinoma (PDAC) disrupts the normal function of the remaining pancreas, leading to altered systemic metabolism and peripheral tissue wasting that begins early in disease progression. Using mouse models of PDAC, we find small pancreas tumors lead to both endocrine and exocrine pancreatic dysfunction that results in systemic nutrient depletion and loss of both muscle and fat tissue. Providing free glucose in the diet that is absorbed despite pancreatic exocrine dysfunction causes hyperglycemia and blunts fat wasting without affecting muscle loss. Muscle mass can be restored by free dietary amino acids or pancreatic enzyme supplementation. Exocrine dysfunction causing reduced dietary protein digestion promotes muscle proteolysis and autophagy. Autophagy is a major driver of muscle wasting in PDAC, as muscle-specific deletion of the core autophagy gene Atg7 also reduces muscle wasting. Disrupting muscle autophagy without restoring systemic nutrition slows tumor growth and improves survival of mice with PDAC. Tracing the fate of amino acids released from muscle of mice with PDAC shows redistribution to both tumor and host tissues. Notably, improving nutrition in mice with disrupted muscle autophagy promotes tumor growth. Together, the data argue that early peripheral tissue wasting associated with early pancreatic cancer is driven by altered normal pancreatic organ function that leads to reduced nutrition and enhanced muscle autophagy, releasing nutrients to support both tumor and host metabolism.

cancer biology↗

bioRxiv2026

Generative single-cell transcriptomics via large language models

Single-cell and spatial transcriptomics have generated vast atlases of cellular states, yet these data are almost exclusively used for analysis rather than generation. Here we introduce the LLM-based model PGL, Portraying Gene Language, a framework that reframes single-cell transcriptomes as a language generation modeling problem. PGL represents each cell as a long sequence of gene-expression tokens and uses a large language model to synthesize complete single-cell RNA-seq profiles from metadata alone, such as tissue and disease context. PGL-generated cells recapitulate dataset-specific transcriptomic structure, align with known cancer subtype biology, and mix coherently with real single-cell datasets. Notably, generated cells can be used as effective references for spatial transcriptomics, enabling accurate cell-type mapping without matched single-cell atlases. By shifting single-cell modeling from representation learning to cell generation, PGL enables virtual cohort construction, hypothesis generation, and reference-on-demand analysis, positioning generative language models as foundational tools for in silico single-cell biology.

bioinformatics↗

bioRxiv2026

New paradigm of CRISPR spacerome for classification of global Escherichia coli lineages

Our research team has recently identified a proof-of-concept for a previously unrecognized spacerome signature in a subset of globally distributed E. coli. Motivated by our initial observation, we pursued a more in-depth CRISPR spacerome analysis, which revealed a novel insight in discrimination of global E. coli sequence type (ST) lineages. We systematically retrieved publicly available E. coli complete genomes, analyzed and compared CRISPR spacerome of globally distributed strains. We found that global strains harbored spacerome that has remained conserved spaceromes for decades across multiple continents. Strains with the conserved spacerome belong to specific ST lineages of clinical importance. In addition, most protospacers were assigned to Gammaproteobacteria and Caudoviricetes. Our findings reveal unexpected long-term conservation of CRISPR spaceromes and their potential as high-resolution markers for E. coli epidemiological tracking.

microbiology↗

bioRxiv2025

The Wnt/APC destruction complex targets SREBP2 in a β-catenin-independent pathway to control cholesterol metabolism

APC, the core scaffold of the Wnt destruction complex, targets the transcriptional co-activator {beta}-catenin for proteolysis. There is no convincing evidence that APC directs degradation of other substrates. Using a reconstituted cytosolic extract-based system and complementary in vivo and cellular assays, we show that SREBP2, the master regulator of cholesterol biosynthesis, is a direct APC-AXIN1 substrate. APC-dependent SREBP2 degradation is conserved in Xenopus embryos, mouse colon, and human colorectal cancer cells and restricts SREBP2 target-gene expression, cholesterol synthesis, and tissue cholesterol levels. Mechanistically, APC and AXIN1 promote SREBP2 degradation via a conserved phosphodegron, which marks SREBP2 for ubiquitination by the E3 enzyme, FBXW7. Like {beta}-catenin, SREBP2 is stabilized by extracellular Wnt ligands; unlike {beta}-catenin, its regulation is independent of GSK3{beta} and CK1 and requires the entire APC mutational cluster region (MCR), whereas {beta}-catenin turnover can operate with only a partial MCR. These findings define a {beta}-catenin-independent branch of Wnt signaling that couples APC to sterol metabolism, providing a mechanistic rationale to target the mevalonate/SREBP2 axis in APC-mutant colorectal cancer.

biochemistry↗

bioRxiv2025

Heterochronic myeloid cell replacement reveals the local brain environment as key driver of microglia aging

Aging, the key risk factor for cognitive decline, impacts the brain in a region-specific manner, with microglia among the most affected cell types. However, it remains unclear whether this is intrinsically mediated or driven by age-related changes in neighboring cells. Here, we describe a scalable, genetically modifiable system for in vivo heterochronic myeloid cell replacement. We find reconstituted myeloid cells adopt region-specific transcriptional, morphological and tiling profiles characteristic of resident microglia. Young donor cells in aged brains rapidly acquired aging phenotypes, particularly in the cerebellum, while old cells in young brains adopted youthful profiles. We identified STAT1-mediated signaling as one axis controlling microglia aging, as STAT1-loss prevented aging trajectories in reconstituted cells. Spatial transcriptomics combined with cell ablation models identified rare natural killer cells as necessary drivers of interferon signaling in aged microglia. These findings establish the local environment, rather than cell-autonomous programming, as a primary driver of microglia aging phenotypes.

neuroscience↗

bioRxiv2025

Angiopoietin signalling is a central axis of amyloid-driven vascular dysfunction in Alzheimer's disease

The neurovascular unit is critical for brain health, and its dysfunction has been linked to Alzheimers disease (AD). However, a cell-type-resolved understanding of how diverse vascular cells become dysfunctional and contribute to disease has been missing. Here, we applied Vessel Isolation and Nuclei Extraction for Sequencing (VINE-seq) to build a comprehensive transcriptomic atlas from 101 individuals along AD progression. Our analysis of over 842,646 parenchymal and vascular nuclei reveals that vascular dysfunction in AD is driven by transcriptional changes rather than shifts in cell proportions, with brain endothelial cells (BECs) and smooth muscle cells (SMCs) most affected. Strikingly, these molecular signatures emerge early at the mild cognitive impairment (MCI) stage, implicating vascular dysfunction early in AD pathogenesis. Stratifying by pathology reveals distinct vascular responses to {beta}-amyloid and tau: {beta}-amyloid burden primarily perturbs BECs and SMCs, while tau pathology predominantly impacts glial cells. We identify dysregulated angiopoietin signaling across multiple vascular cell types as a key axis, with antagonistic ANGPT2 in vascular cells and ANGPT1 in astrocytes becoming progressively dysregulated with AD. Together, this work provides a foundational resource that reveals early and pathology-specific pathways of vascular dysfunction in AD. Key MessagesO_LIVINE-seq analysis from 101 individuals creates a comprehensive human brain vascular atlas across Alzheimers disease (AD) progression. C_LIO_LIAD vascular dysfunction is driven by transcriptional changes rather than shifts in cell proportions, with BECs and SMCs most affected. C_LIO_LITranscriptional signatures of vascular dysfunction emerge early at the mild cognitive impairment (MCI) stage, preceding severe cognitive symptoms and aligning more closely with AD than cognitively normal individuals. C_LIO_LIA{beta} and tau associate with distinct vascular changes: A{beta} mainly perturbs endothelial and smooth muscle cells, while tau impacts microglia and astrocytes. C_LIO_LIAngiopoietin signaling (antagonistic ANGPT2 in vascular cells vs. ANGPT1 in astrocytes) becomes progressively dysregulated during AD progression. C_LI

neuroscience↗

bioRxiv2025

Field evaluation of non-invasive sampling for monitoring nine swine pathogens in the Republic of Korea

BackgroundNon-invasive sampling methods, such as swine oral fluid (OF) and fecal boot (FB) sampling, have gained attention due to their animal welfare considerations, ease of collection, and cost effectiveness. However, comparative data regarding non-invasive and standard sampling methods remain limited, particularly in the Republic of Korea, where research on OF and FB sampling is sparse. Therefore, we investigated representative pig pathogens, including six porcine respiratory pathogens and three gastrointestinal pathogens, using OF and FB samples for pathogen surveillance across commercial farms in the Republic of Korea. MethodsThe results from targeted polymerase chain reaction (PCR) assays for major respiratory and gastrointestinal pathogens obtained using OF and FB samples were compared with those obtained from standard samples. Non-invasive samples included 30 OF ropes and 30 FBs collected from nine farms, and 360 standard samples (blood, fecal, oral, and nasal swabs) were also collected. PCR assays for the nine pathogens were conducted, and the results from OF, FB, and combined OF+FB were compared with those from standard samples at the pen level. ResultsOF sampling showed superior test concordance with respiratory pathogens than gastrointestinal pathogens, and FB sampling demonstrated superior detection of gastrointestinal pathogens compared with OF sampling. Additionally, the combination of OF+FB sampling improved the detection concordance for all nine target pathogens investigated. ConclusionsWe demonstrated the feasibility and potential applicability of OF and FB sampling under field conditions in Korean pig farms. Therefore, this study supports OF and FB sampling as useful tools for swine pathogen monitoring in both research and practical field applications.

microbiology↗

bioRxiv2025

Feature-specific inhibitory connectivity augments the accuracy of cortical representations

To interpret complex sensory scenes, animals exploit statistical regularities to infer missing features and suppress redundant or ambiguous information. Cortical microcircuits might contribute to this cognitive goal by either completing or cancelling predictable activity, but it remains unknown whether, and how, a single circuit can implement these antagonistic computations. To address this central question, we used all-optical physiology to simulate sensory-evoked activity patterns in pyramidal cells (PCs) and somatostatin interneurons (SSTs) in the mouses primary visual cortex (V1). In the absence of external visual input, photostimulation of orientation-tuned PC ensembles drove either completion or cancelation of input-matching representations, depending on the number of photostimulated cells. This dual computational capacity arose from the co-existence of like-to-like excitatory interactions between PCs, and a newly discovered like-to-like SST-PC connectivity motif, in which SSTs are preferentially recruited by, and in turn suppress, similarly tuned PCs. Finally, we show that photoactivation of tuned SST ensembles during visual processing improved the discriminability of their preferred visual input by suppressing ambiguous activity. Thus, these complementary feature-specific connectivity motifs allow different strategies of contextual modulation to optimize inference by either completion (through PC-PC interactions) or cancelation (via PC-SST-PC loops) of predictable activity, depending on the structure of the input and the network state.

neuroscience↗

bioRxiv2025

Targeting RUNX1 in Macrophages Facilitates Cardiac Recovery

Despite advances in disease treatment, our understanding of how damaged organs recover and the mechanisms governing this process remain poorly defined. Here, we mapped the transcriptional and regulatory landscape of human cardiac recovery using single cell multiomics. Macrophages emerged as the most reprogrammed cell type. Deep learning identified the transcription factor RUNX1 as a key regulator of this process. Macrophage-specific Runx1 deletion recapitulated the human cardiac recovery phenotype in a chronic heart failure model. Runx1 deletion reprogrammed macrophages to a reparative phenotype, reduced fibrosis, and promoted cardiomyocyte adaptation. RUNX1 chromatin profiling revealed a conserved regulon that diminished during recovery. Mechanistically, the epigenetic reader BRD4 controlled Runx1 expression in macrophages. Chromatin activity mapping, combined with CRISPR perturbations, identified the precise regulatory element governing Runx1 expression. Therapeutically, small molecule Runx1 inhibition was sufficient to promote cardiac recovery. Our findings uncover a druggable RUNX1 epigenetic mechanism that orchestrates recovery of heart function.

immunology↗

bioRxiv2025

Neural variability structure in primary visual cortex is optimal for robust representation of visual similarity

How neuronal populations construct robust representations of the sensory world despite neural variability remains unclear. Here, we show that trial-to-trial variability in mouse primary visual cortex follows a simple rule: for each stimulus, the mean and variance of spike counts across neurons show a highly stereotyped relationship, with a slope of 1 on a log-log scale. To test how this geometry of trial-to-trial variability affects sensory representations, we numerically manipulated the slope of the log-mean vs. log-variance relationship. We found that the intrinsic geometry of trial-to-trial variability, with slope 1, enables representations of distinct sensory inputs to have minimal overlap while being continuous. At this slope, representational similarity was maximally consistent across neuronal subsets, both within and across mice. By contrast, at slope 0, when variance is uniform across neurons, visual representations were more efficient but less robust. Simulations of recurrent networks, where excitatory and inhibitory neurons formed locally balanced clusters with stronger connections, showed that larger clusters produced steeper slopes and more robust representations. Together, these results suggest that the geometry of trial-to-trial variability mediates a tradeoff between efficiency and robustness, and that the intrinsic geometry is near optimal for robust visual representations.

neuroscience↗

bioRxiv2025

Enhanced anti-nociception by novel dual antagonists for 5HT2aR and mGluR5 in preclinical models of pain

Significant research has focused on developing anti-nociceptive pain treatments by targeting specific molecular candidates. The serotonin 2A receptor (5-HT2AR) and metabotropic glutamate receptor 5 (mGluR5) are recognized as key mediators in neuropathic pain. However, the combined effects of simultaneous inhibition of these targets remain unexplored. This current study investigated the therapeutic potential of concurrently antagonizing 5-HT2AR and mGluR5. Using spinal nerve ligation (SNL) and formalin-induced pain models in male Sprague-Dawley rats, we demonstrated that the simultaneous administration of both antagonists significantly enhanced anti-allodynic and anti-nociceptive effects, resulting in increased allodynia thresholds and reduced pain-related behaviors. This dual antagonism provided pain relief comparable to that of gabapentin and morphine. Furthermore, novel small molecules designed to simultaneously antagonize 5-HT2AR and mGluR5 exerted anti-nociceptive effects by suppressing excitatory postsynaptic responses and inhibiting the phosphorylation of ERK and AKT signaling molecules. Notably, the dual antagonist maintained anti-allodynic efficacy with repeated administration, unlike morphine, which exhibited clear tolerance development with daily use. Moreover, when administered intravenously, the dual antagonist demonstrated a low potential for abuse. These findings indicate that the simultaneous antagonism of 5-HT2AR and mGluR5 represents a promising pharmacological target for the management of chronic pain. This approach may offer enhanced analgesic outcomes while reducing the risk of undesirable side effects, such as tolerance and the potential for abuse.

neuroscience↗