搜索结果：computational

Access to large, diverse biomedical datasets is critical for advancing medical research, yet privacy regulations severely restrict data sharing. We present an end-to-end framework for privacy-preserving health data synthesis that integrates advanced deep generative models (DGMs) with robust preprocessing, formal differential privacy (DP) training for select DGMs, empirical privacy risk evaluation, data-sufficiency analysis, domain-guided quality control, and biobank visualization tools. Released as open-source containerized software, the framework ensures reproducible deployment while preserving statistical fidelity, machine learning (ML) utility, and privacy guarantees. Empirical evaluations across diverse biobank datasets demonstrate that TabSyn-a transformer-based diffusion model-combined with our correlation-and distribution-aware CorrDst loss function achieves superior performance balancing fidelity, privacy, and computational efficiency. The tailored preprocessing pipeline effectively handles high missingness rates, substantially improving distributional accuracy and clinical plausibility. Across 26 biobank datasets spanning three regulatory levels, the framework shows that TabSyn with correlation- and distribution-aware loss function consistently achieves superior performance in terms of fidelity, privacy, and computational efficiency.

Reliable detection of bolt loosening in safety-critical infrastructure requires monitoring that captures microsecond-scale transients. However, processing 1 MHz vibration signals at the edge presents a fundamental dilemma: standard downsampling can smear sparse diagnostic impulses, whereas full-bandwidth processing is often computationally prohibitive. Here we propose PGRF-Net, a physics-guided deep learning framework that reconciles transient preservation with extreme compression. PGRF-Net integrates (1) a physics-guided resampling operator for high-rate vibration signals (e.g., 150 : 1 compression); (2) a heterogeneous gradient-spectral tensor representation that augments time-frequency information with morphology-aware channels; and (3) an asymmetric fusion module over two representations derived from the same signal (pseudo-image representation + waveform representation). On a six-class 1 MHz PVDF vibration dataset (109,668 segments; temporal-split test = 16,131), PGRF-Net reaches a best-run clean test accuracy of 95.12% under a block-wise temporal split. A strict file-disjoint split is also reported as a cross-scene hold-out protocol. On an independent Zenodo benchmark (3-fold file-disjoint CV, 9 runs per experiment), we evaluate the proposed pipeline together with three feature-engineering controls (Exp 501-503). These results support a practical compression-learning pipeline for industrial monitoring where both transient fidelity and computational efficiency are required.

Ovarian cancer (OC) remains a malignancy characterized by obscure risk factors and unfavorable prognosis. While 3-tert-butyl-4-hydroxyanisole (3-BHA) is suspected of exerting toxic effects on ovarian health, the precise molecular mechanisms underlying its impact remain elucidated. This study aims to systematically investigate the potential pathogenic mechanisms of 3-BHA in the progression of OC.Integrated transcriptomic data from the GEO database (GSE18520 and GSE40595) were analyzed. A synergistic computational framework was employed, incorporating Differentially Expressed Genes (DEGs) identification, Weighted Gene Co-expression Network Analysis (WGCNA), multiple machine learning algorithms, and SHapley Additive exPlanations (SHAP) analysis to achieve high-interpretability feature selection.Five hub genes-CXCR4, CCL7, CXCL8, CXCR2, and CX3CL1-were identified, all demonstrating robust diagnostic efficacy with AUC values of 0.911, 0.882, 0.823, 0.772, and 0.837, respectively. Prognostic profiling via GEPIA3 highlighted CXCR2 overexpression as a potential critical biomarker driving poor clinical outcomes in OC. Furthermore, molecular docking validated the strong binding affinity of 3-BHA with CX3CL1 and CXCR2. Subsequent 100 ns molecular dynamics simulations and thermodynamic stability assessments confirmed the structural stability of the 3-BHA-CXCR2 complex.By integrating bioinformatics and computational toxicology, this study deciphers the potential mechanistic landscape through which 3-BHA influences OC. These findings not only refine the toxicological understanding of 3-BHA but also provide novel candidates for early diagnosis and prognostic risk stratification in OC.

Deep learning models for medical image analysis often rely on large-scale parameterization, which may limit their practical use in resource-constrained settings. This study aims to design a structurally compact multi-source framework capable of delivering competitive diagnostic performance with reduced computational overhead. We propose ML-ConvNet, a lightweight architecture comprising approximately 4.2 K parameters and 924 M FLOPs at 512×512 input resolution. The network incorporates Multi-Branch Re-parameterized Convolutions for scale-aware feature extraction, Hierarchical Dual-Path Attention for feature localization, Feature Self- Transformation for cross-feature interaction, and a Local Variance Weighted optimization strategy to address class imbalance. The framework is evaluated independently on three publicly available benchmark datasets representing heterogeneous imaging modalities: brain MRI, lung CT, and chest X-ray. Ablation studies, precision-recall analysis, cross-modality validation, and computational benchmarking are conducted to assess performance, stability, and efficiency under controlled experimental conditions. Within the evaluated settings, results indicate competitive diagnostic accuracy relative to established lightweight baselines, including EfficientNet and MobileNet variants, while substantially reducing parameter count. Class-wise F1-scores and PR-AUC values suggest relatively stable minority-class performance under repeated cross-validation sampling. Attention visualizations show activations concentrated over regions broadly associated with pathological findings, though these observations are qualitative in nature. Inference latency measurements on CPU and mobile hardware suggest feasibility for low-latency deployment under the tested single-image batch configurations, though real-world throughput may differ depending on hardware and operational conditions. These findings suggest that careful architectural design and domain-informed inductive biases may support competitive medical image classification on public benchmark datasets without extensive parameter scaling. The framework was evaluated exclusively under controlled conditions on publicly available data, and multi-institutional external validation is required before conclusions regarding generalizability or clinical applicability can be drawn.

Aiming at the control problems of strong nonlinearity, multi-channel coupling and complex aerodynamic disturbance of flapping-wing aircraft, this study proposes an environment-adaptive cascade control switching method based on the PID architecture. A physical control closed-loop is established using MWORKS.Sysplorer, with classical PID as the outer main loop and PID, nonlinear compensation PID, and SMC as the inner loop to form a nested structure, which is cross-validated via MATLAB/Simulink. Through the disturbance quantization parameter k, operating conditions are classified into high, medium, and low disturbances for strategy switching. At the classical flapping frequency of 15 Hz, the proposed method converges the attitude error to within 0.06 rad, improves the anti-disturbance performance under high disturbance by 46.2% compared with traditional PID, constrains the phase lag to the stable interval corresponding to the natural frequency of 50 rad/s, and optimizes computational efficiency with the average single-step simulation time ≤ 0.02 s. This method addresses the insufficient full-working-condition adaptability of traditional single control strategies, and provides a highly robust implementation approach for the control of flapping-wing aircraft under complex disturbances.

Ectonucleotidases, including NTPDases and ecto-5'-nucleotidase (e-5'NT/CD73), regulate extracellular purinergic signaling by converting ATP to adenosine, a pathway critically involved in immune response, inflammation, and cancer progression. In this study, a novel library of 22 N-propylsulfonyl-substituted indole-based hydrazinecarbothioamides (5a-5v) was synthesized and structurally characterized. Biological evaluation against human e-5'NT and NTPDase1, -2, -3, and - 8 revealed that several compounds exhibited low micromolar inhibitory activity, with 5n (IC50 = 1.7 µM), 5o (IC50 = 1.7 µM), 5f (IC50 = 1.0 µM), and 5i (IC50 = 1.6 µM) emerging as the most promising derivatives, showing strong potency and isoform selectivity. Structure-activity relationship analysis indicated that both electronic and steric features of substituents significantly influence activity and enzyme preference. Molecular docking studies performed on e-5'NT demonstrated that active compounds adopt consistent binding modes within the catalytic pocket, stabilized by key residues such as Asp-506, Phe-500, Phe-417 and Arg-395. Binding free energy calculations (MM-GBSA) supported strong ligand-protein interactions ( ~ - 70 kcal/mol). The docking protocol was validated by redocking, yielding an RMSD value well below the accepted threshold. Molecular dynamics simulations (500 ns) confirmed stable complex formation, with low RMSD values (~ 1-3 Å), limited residue fluctuations, and persistent interactions with catalytic residues. Surface and compactness parameters (rGyr, SASA) remained stable, indicating consistent ligand accommodation. In silico ADME analysis suggested favorable drug-like properties for most compounds, particularly for the lead candidates. Overall, these findings identify 5n and 5o as the most promising lead compounds, supported by both experimental and computational results, and highlight this scaffold as a valuable platform for the development of selective ectonucleotidase inhibitors.

The PE_PGRS gene family in Mycobacterium tuberculosis exhibits extensive sequence variability across genotypes, which is consistent with antigenic divergence. Here we investigate how Mtb-despite lacking horizontal gene transfer-balances genomic stability with adaptive plasticity. Comparative analysis of 88 bacterial genomes reveals that PE_PGRS genes exhibit features facilitating mutability, including a significantly elevated CGGC tetramer density (mean 4.97 per 100 nt; range 1.7-7.4) compared with the genome-wide average (1.62 per 100 nt; p = 0.011) and depleted in out-of-frame stop codons, potentially conferring robustness to 1-nt and 2-nt frameshifts. Computational predictions suggest that CGGC motifs may promote secondary DNA structures, potentially destabilizing replication and contributing to replication errors, while the scarcity of out-of-frame stop codons allows continued translation beyond frameshifts, leading to changes in protein sequence and length. This dual organization may contribute to the observed adaptability of M. tuberculosis and could highlight a broader principle by which some pathogens evolve under strong constraints on horizontal gene transfer. We propose that CGGC-rich regions may function as programmed mutational hotspots across a wide range of microorganisms.

Cis-regulatory elements (CREs) drive tissue- and cell-specific gene expression and are essential for safe, sustainable genetic control strategies in pest and vector insects, including the engineering of gene drives in the primary human-malaria vector Anopheles gambiae. Yet CREs remain poorly defined in mosquitoes due to limited computational tools and practical methods for identification and validation. We present a systematic in silico approach for CRE discovery, correlating targeted DNA-motif searches with gene expression, followed by frequency and distribution analysis within putative promoter regions. Applied to the A. gambiae germline, this approach identified hundreds of putative CREs significantly correlated with germline expression in one or both sexes, often linked to distinct sperm developmental stages and chromosomal locations, suggesting roles in broader regulatory mechanisms such as dosage compensation and meiotic silencing. When mapped onto pre-characterised germline promoters, CRE distribution aligned with regions associated with experimental expression patterns. Finally, we validated a top-ranked testis-enriched CRE using an in vivo dual-reporter assay, showing that mutation of conserved nucleotides drastically altered male germline expression. To the best of our knowledge this work provides the first nucleotide-resolution regulatory genome annotation of the A. gambiae germline, offering a transferable framework to aid promoter design for genetic control strategies against malaria mosquitoes and other insect pests.

Insurance fraud detection remains challenging to predict in reality because claims data is often uneven among classes, and the information concerning claims is often multidimensional and nonhomogeneous. The present research used a unified evaluation framework to assess the predictive and interpretive capabilities of three distinct model families: CatBoost (tree-based ensemble learning), Bi-GRU with Attention (sequence-oriented learning), and TabTransformer (categorical feature contextual). The model families were tested using a standardised experimental protocol.The study is novel in the sense of a cross-model interpretability framework that unites Shapley Additive Explanation (SHAP)-based feature attribution with attention-based contextual analysis to enable a clear comparison of model reasoning between the suggested frameworks. The data on which the experiments were done consisted of 4,000 life insurance claims that were characterized in terms of 83 attributes. Common preprocessing procedures like missing values, scaling numerical variables, and selecting highly correlated variables were used before training the models. Experimentally, CatBoost is proven to be the most precise on legitimate claims, Bi-GRU is the most recall on fraudulent claims, and TabTransformer is the best in terms of tradeoff between accuracy, interpretability, and computational efficiency. Practical characteristics such as the quantity of claim, tenure in a policy, and diagnosis were repeatedly emphasized in both SHAP and attention analyses. Combined, the current research study provides a consistent and explainable benchmark that may be applied to conduct fraud detection research reliably and assist practitioners in choosing models that are accurate and understandable.

Parkinson's Disease (PD) is a progressive neurodegenerative disorder that causes motor and cognitive impairments, affecting approximately 1% of individuals over 60 years of age. Speech impairments are among the earliest and most accessible biomarkers, making voice-based assessment a promising avenue for remote PD monitoring. However, existing speech-based PD prediction methods suffer from feature redundancy that degrades model performance, non-Gaussian data distributions that violate model assumptions, and limited systematic feature grouping strategies. This study introduces an adaptive approach to improve PD diagnostic precision by predicting the Motor Unified PD Rating Scale (UPDRS) and Total-UPDRS scores from biomedical voice measurements. The proposed framework addresses these challenges through three integrated components: (1) Box-Cox transformation to stabilize variance, reduce skewness, and normalize features; (2) a clustering-based feature selection method that groups correlated features via K-Means and selects the most informative representative per cluster using mutual information, thereby eliminating redundancy without losing discriminative power; and (3) an Extra Trees Regressor (ETR) whose extreme randomization in node splitting provides computational efficiency and reduced variance. To ensure rigorous evaluation, a subject-independent data splitting strategy is adopted to prevent data leakage, and k-fold cross-validation is employed to assess model stability. The proposed method is compared against multiple feature selection techniques-mutual information, recursive feature elimination, Lasso regression, and autoencoders-paired with nine regression models including Ridge, Lasso, Linear, Decision Tree, k-Nearest Neighbors, Random Forest, Gradient Boosting, AdaBoost, and Extra Trees Regressors. The clustering-based feature selection combined with ETR yielded the best performance, achieving [Formula: see text] scores of 0.999 for Motor-UPDRS and 0.997 for Total-UPDRS on the test set. These results are further supported by cross-validation analysis and feature importance evaluation, demonstrating the effectiveness and robustness of the proposed framework for speech-based PD telemonitoring.

RNA modifications regulate post transcriptional gene expression, yet most computational methods model each modification independently and overlook competition among modification types at a single site. We present EvoRMD, a biologically contextualized and interpretable framework for RNA modification prediction. EvoRMD combines RNA language model embeddings with structured metadata, including species, organ, cell type, and subcellular localization, and uses attention to identify informative sequence positions. A shared multiclass classifier produces context conditioned predictions across 11 modification types. EvoRMD achieves strong performance and provides interpretable insights through attention patterns and motif analyses, supporting biologically grounded prioritization of candidate RNA modifications.

The human visual system can estimate the three-dimensional shapes of translucent objects. However, the shape estimation of translucent objects is less accurate than that of opaque objects. The specular component is particularly important in the shape perception of translucent objects because it is robust, whereas the non-specular component is affected by translucency. Previously, we developed a shape recovery algorithm as a computational model of human shape perception from opaque specular images. The algorithm primarily uses the specular component and secondarily uses the non-specular component. In the current study, the shape recovery performance of this algorithm for translucent objects was evaluated against ground-truth shapes. The results showed that the reconstruction of the three-dimensional shapes of low-translucency objects by the algorithm was comparable to that of opaque specular images. A modification of the algorithm involving the reversal of the non-specular component was effective for high-translucency objects. However, even with this modification, the estimation accuracy for high-translucency objects was lower than that for opaque objects. These results provide new insight into the possible mechanisms of shape perception for translucent objects.

In this study, a novel series of chromene-based derivatives was rationally designed as potential VEGFR-2 inhibitors based on key structural and pharmacophoric features required for antiangiogenic activity. Accordingly, twelve chromene derivatives (13a-e, 15a-e, and 17a-b) were successfully synthesized and structurally characterized. The synthesized compounds were evaluated in vitro for their cytotoxic activity against human cancer cell lines (MCF-7, HepG-2, and HCT-116), in addition to normal WI-38 and WISH cells. Among the tested compounds, compound 13a demonstrated the most potent and selective antiproliferative activity, exhibiting low micromolar IC50 values and favorable selectivity indices. Enzymatic assays confirmed its VEGFR-2 inhibitory activity (IC50 = 1.666 ± 0.025 µM), comparable to the reference drug sorafenib. Mechanistic investigations revealed that compound 13a effectively inhibited cancer cell migration in a wound healing assay, highlighting its potential antiangiogenic properties. Furthermore, compound 13a induced significant G0/G1 cell cycle arrest in MCF-7 cells and triggered apoptosis, as evidenced by Annexin V/PI staining. To support the experimental findings, Density Functional Theory (DFT) calculations confirmed favorable structural stability and electronic properties. Molecular docking studies demonstrated strong binding interactions within the VEGFR-2 ATP-binding site. These results were further validated by 200 ns molecular dynamics simulations, MM-GBSA binding free energy calculations, Protein-Ligand Interaction Fingerprints (Pro-LIF), Principal Component Analysis of Trajectories (PCA-T), and Free Energy Landscape (FEL) analyses, confirming the dynamic stability and favorable energetics of the VEGFR-2-13a complex. Overall, this integrated experimental and computational study identifies compound 13a as a promising VEGFR-2-targeted anticancer lead warranting further preclinical investigation.

As an extension of standard graphs, hypergraphs have demonstrated significant advantages in modeling high-order complex relationships compared with standard graphs. Existing literature has witnessed the great success of hypergraph representation learning methods in classifying nodes. However, most of them seek to obtain low-dimensional crisp representations, overlooking the fuzzy and uncertain nature of node attributes. In fact, node attributes such as paper keywords may contain noise or be incomplete, which leads to uncertain semantics. To address this issue, in this paper, we propose learning fuzzy representations for hypergraph node classification. Specifically, we develop a novel method called Hypergraph Collaborative Fuzzy Network (HyperCFN), which studies hypergraph representations with fuzzy logic. Firstly, HyperCFN augments the original hypergraph into two hypergraphs, which are then put into the proposed fuzzy hypergraph encoders. The fuzzy hypergraph encoders consist of hypergraph collaborative networks and fuzzy logic to learn fuzzy representations for every node and hyperedge. Subsequently, the learned representations are enforced node-, hyperedge-, and membership-level contrast. Lastly, to further preserve the hypergraph structure, we develop decoders to reconstruct the augmented hypergraphs. We perform extensive experiments on several datasets, and the promising results demonstrate that the effectiveness of the proposed model and learning fuzzy representations for hypergraphs is valid.

Expanders in organic Rankine cycle systems serve as critical energy-conversion components in low-grade waste heat recovery installations, yet their reliable operation is threatened by faults such as bearing defects, rotor imbalance, and blade cracking. Conventional diagnostic methods often struggle with non-stationary vibration characteristics, class imbalance, and low signal-to-noise ratios inherent to these working environments. This paper proposes an improved deep residual network, referred to as multi-scale convolutional block attention module residual network, that integrates a multi-scale parallel feature extraction module with convolutional block attention mechanisms for intelligent fault diagnosis. The multi-scale module employs three parallel convolutional branches with different kernel sizes to simultaneously capture transient impulses, periodic modulation, and low-frequency envelope features across multiple temporal scales. Attention-enhanced residual blocks sequentially recalibrate channel and spatial responses to emphasize fault-sensitive features while suppressing noise interference. A training optimization scheme combining Focal Loss, cosine annealing, and targeted data augmentation is further introduced to address the small-sample imbalanced-data challenge. Five-fold cross-validation experiments conducted on a 10 kW single-screw expander test rig demonstrate that the proposed model achieves 98.11 ± 0.34% diagnostic accuracy across four health states, surpassing the standard deep residual network baseline by 6.57 percentage points, with only 3.27% relative accuracy degradation at 10 dB signal-to-noise ratio. Ablation studies confirm a multiplicative synergy between the multi-scale and attention modules, statistical significance tests validate the robustness of the observed improvements, and comparative evaluations against six benchmark methods demonstrate the superiority and generalizability of the proposed approach.

Proton pump inhibitor (PPI) use has been associated with metabolic dysfunction associated with steatotic liver disease (MASLD) in multiple studies. While the association is confounded by various risk factors, such as BMI and age, a potential mediating factor of the microbiome has been suggested. In this study, we aimed to identify bacterial clades with the highest mediating potential and evaluate the serially mediated path through microbially derived endogenous ethanol. Microbiome mediation analysis of PPI use and MASLD was conducted in two cohorts. In a bariatric surgery cohort (n = 122), liver biopsy-proven steatosis grade and postprandial ethanol concentrations were used as outcomes. In the HELIUS cohort (n = 2440), a general population cohort study, mediation was performed using the Fatty Liver Index (FLI) score. The strongest associations were validated in the FINRISK cohort (n = 7066). Several bacterial taxa, which are predominantly found in the small intestine, showed a potential role in mediating the effects of PPIs on MASLD, postprandial ethanol levels, and FLI score. The Lactobacillales order showed the strongest mediating potential across the outcomes tested in both discovery cohorts. A notable serial mediation pathway was identified, linking PPI use to MASLD via Lactobacillales abundance and postprandial plasma ethanol concentrations. The mediating role of Lactobacillales in the association between PPI use and FLI scores was confirmed in the final study cohort. Data from multiple cross-sectional cohort studies support a mediating potential of the microbiome in the association between PPI use and hepatic steatosis, independent of alcohol consumption. The effect of PPIs on MASLD appears to be mediated mainly by increased lactic acid bacteria abundance, and is potentially, in part, serially mediated by endogenous ethanol production.

Long-tail segmentation is a crucial challenge in computer vision, where most models prioritize common head classes over rare tail classes. This problem is particularly prominent in retinal vessel segmentation, as conventional approaches often struggle to overcome underrepresented faint vessels, noise-induced boundary ambiguity, and excessive parameters that prohibit portable deployment. To address these challenges, we introduce LFU-Net, a lightweight and clinically applicable method for long-tail retinal vessel segmentation. It integrates a three-component ensemble: a Frequency-Aware Encoder with a Multi-Branch Frequency Convolution block, which uses wavelet decomposition to suppress noise and retain details; Hierarchical frequency-token enhanced Low-Rank Adaptation, which efficiently enhances the representation of tail classes (faint vessels) with minimal parameters; and a Recursive Residual Attention Fusion module to ensure vascular topological continuity. Extensive experiments on four public benchmark datasets demonstrate that LFU-Net achieves competitive performance compared to recent relevant models. Its lightweight nature supports real-time inference on portable devices. Ablation studies confirm the improvement contribution of each core component, indicating its potential utility in early disease detection when clinical resources are limited.

A position paper released by the European Association of Nuclear Medicine emphasised the need for multidisciplinary engagement to establish dosimetry-based personalised treatment in Radionuclide therapy (RNT). The uncertainty analysis results often ignored in routine clinical practice should be incorporated into the dose calculations to improve the efficacy and accuracy of treatment. In this study, patients with haematological malignancies undergoing radioimmunotherapy were evaluated. Our study aimed to calculate the uncertainties associated with each parameter of the single time point (STP) dosimetry chain and compare the with multiple time points (MTP) in the bone marrow and liver results. 28 patients received an intravenous injection of 111In-besilesomab (0.17 ± 0.01GBq) for pre-therapeutic dosimetry and were subsequently treated with 90Y-besilesomab(2.43 ± 0.53GBq). A dosimetry analysis was performed on bone marrow (BM) and liver with MTP and STP. We investigated the uncertainty in population mean effective half-life, volume, recovery coefficient, counts, measured activity, fitting parameters, time-integrated-activity, S-factors, and absorbed dose (AD) for a group of patients. The mean absorbed dose per unit administered activity (DpA) to BM was 5.8 ± 1.7 mGy/MBq with MTP and 5.8 ± 1.6 mGy/MBq with STP, and to the liver was 2.9 ± 1.9 mGy/MBq with MTP and 3.1 ± 2.4 mGy/MBq with STP. The mean fractional uncertainty associated with total absorbed dose to BM was 13.18 ± 3.46% with MTP and 18.75 ± 3.22% with STP, and to liver was 5.77 ± 3.13% with MTP and 49.78 ± 25.36% with STP. A moderate positive relationship (R2 = 0.7) was noted between post-injection acquisition time and AD uncertainty with STP for BM, whereas a strong positive relationship (R2 = 1) was noted for the liver. The absorbed dose uncertainty in STP was significantly higher compared to the MTP. Incorporating the uncertainty analysis for STP dosimetry parameters in routine clinical practice is strongly recommended. The accuracy in the acquisition time, population-based half-life and fitting function for time activity curve is vital for minimising uncertainty in STP dosimetry, which is less time-consuming and easier to implement in clinical practice than MTP.

Dried blood spot (DBS) biosampling holds promise for expanding routine viral load (VL) monitoring for youth with HIV (YWH), particularly those at highest risk for HIV medication non-adherence. This mixed methods study piloted home-based DBS collection with YWH, aged 15-24 years. We enrolled 34 YWH with suppressed VL from April 22, 2020, to December 15, 2021, a subset of a fully virtual, nationwide decentralized clinical trial (ATN 144 SMART). Participants were mailed a HemaSpot™-HF kit and asked to complete a computer-assisted self-interview (CASI) with an instructional DBS video. Surveys and semi-structured interviews provided quantitative and qualitative data to assess feasibility, appropriateness, and acceptability of home-based DBS for VL monitoring. Of 239 total screener attempts, 134 individuals were eligible and 115 provided contact information/completed the screener; 34 enrolled and returned DBS kits. Descriptive analyses showed a positive relationship between perceived suitability, feasibility, and acceptability. Perceived suitability was negatively associated with age, and feasibility differed significantly by health insurance coverage. Qualitative findings identified facilitators such as clinic/provider support, awareness of DBS innovation, insurance coverage, and streamlined mailing processes. Barriers included living environment challenges, cost concerns, and mail delivery issues. This pilot supports a self-management model and provides preliminary evidence that home-based DBS collection is feasible and acceptable among YWH. Scaling up this method through clinic and provider promotion could transform YWH HIV care by enabling remote VL monitoring. Findings also underscore the value of DBS as a practical biospecimen collection strategy for decentralized research models.