Biology subjects

Kuang, N.

Publications and source records attributed to Kuang, N..

4 recordsLinked to original sources

bioRxiv2026

Comorbid HIV and Cocaine Use Exacerbate Accelerated Brain Aging

BackgroundHIV and cocaine use (CU) each relate to cognitive deficits and brain abnormalities, yet their combined impact on brain aging remains unclear. This study examined how comorbid HIV and CU relate to brain aging and cognitive impairment. MethodsWe trained a morphometry-based brain-age model using harmonized Human Connectome Project-Aging data (HCP-A; n=725) with Gaussian Process Regression. The model was applied to an independent cohort with varying HIV/CU burden (HIV-/CU-, n=34; one disorder [HIV+/CU- or HIV-/CU+], n=72; HIV+/CU+, n=80). Brain age gap (BAG; predicted minus chronological age) was examined in relation to comorbidity burden and neurocognitive impairment (NCI; NIH Toolbox), adjusting for age, sex, education, depression, and image-quality indices. Analyses on SHapley Additive exPlanation (SHAP) values characterized network-wise feature-level contributions to brain age estimates. ResultsA dose-dependent effect of comorbidity burden on BAG was observed, with the HIV+/CU+ group showing the highest BAG. Greater BAG was associated with increased likelihood of NCI, and BAG partially mediated the relationship between comorbidity burden and NCI, with a stronger mediation effect in the two-disorder group than in the one-disorder group. Structural contributors to elevated BAG in the HIV/CU cohort included cortical thickness in the visual, ventral attention, and frontoparietal networks, and sulcal depth in the sensorimotor network. ConclusionComorbid HIV/CU is linked to accelerated structural brain aging. BAG may reflect brain-level alterations underlying the association between comorbid HIV/CU and cognitive impairment, and may help identify network-specific targets for intervention.

neuroscience↗

bioRxiv2026

Plasma proteomics reveal heterogeneous subtypes of depression linked to inflammation and aging

The clinical heterogeneity of depression has defied biological classification, limiting personalized treatment. Previous neuroimaging- or symptom-based subtyping of depression failed to clarify the underlying pathoetiology, while plasma proteins which integrates signals from multiple organ systems, offers a promising way to define biologically grounded subtypes. Using plasma proteomics from 2,127 incident depression cases in a cohort of 53000 individuals, we identified three biologically distinct subtypes differing in inflammation, aging, and metabolic profiles. The most prevalent subtype ( inflammation/ageing) was characterized by aging-related inflammation, poorest prognosis with hippocampal atrophy and highest suicide risk, mediated by age-related amygdala atrophy; this subtype had highest anhedonia burden. A distinct inflammation/energy dysregulation group had metabolic pathway enrichment with high inflammation and lifestyle risk factors (smoking) but no ageing trend, predominantly physical/psychomotor symptoms and decreased thalamic volume. In contrast, the inflammation-resilient group had the lowest inflammatory proteomic loading, lowest depression severity, more resilient lifestyle and increased hippocampal volume. These proteomic signatures, detectable years before symptom onset, enable risk stratification and suggest subtype-specific targeted physical and lifestyle interventions.

neuroscience↗

bioRxiv2026

Disrupted dynamics of brain structure function coupling link genetic risk of Alzheimer's Disease and aging to cognitive decline in 34,067 adults

AbstractUnderstanding how a stable structural connectome supports flexible cognition, especially in aging, is a fundamental question in neuroscience. While static structure-function coupling (SFC) is well-studied, the dynamic decoupling of functional activity from structural constraints--dynamic SFC (DSFC)--remains poorly understood. Leveraging MRI data from 34,067 UK Biobank participants (ages 45-82), we characterized the distinct roles of SFC and DSFC in aging, cognition, and health. We found that SFC and DSFC followed spatially divergent aging trajectories: SFC declined primarily in sensorimotor systems, whereas DSFC decreased most prominently in higher-order networks. Both SFC and DSFC in higher-order networks were positively correlated with cognitive performance (e.g., fluid intelligence). However, the associations with mental and physical health diverged between the two measures: reduced DSFC was predominantly linked to health burdens in high-order default/limbic networks, whereas weakened SFC was primarily associated with health burdens in low-order sensory-motor networks. Genetic analyses revealed that Alzheimers risk, specifically APOE {varepsilon}4 dosage, significantly reduced DSFC in higher-order cognitive networks and SFC in visual cortex. Mediation analyses further demonstrated that aging and APOE {varepsilon}4-linked cognitive decline were mediated via visual SFC and ventral attention DSFC. These findings position static and dynamic coupling as complementary mechanisms: static SFC preserves network robustness, while dynamic SFC enables transient reconfiguration for complex integration. Together, our results highlight these dual mechanisms as crucial for maintaining cognitive flexibility, providing potential biomarkers for age-related neurodegeneration.

neuroscience↗

bioRxiv2022

Efficient End-to-end Learning for Cell Segmentation with Machine Generated Incomplete Annotations

Automated cell segmentation from optical microscopy images is usually the first step in the pipeline of single-cell analysis. Recently, deep-learning based algorithms have shown superior performances for the cell segmentation tasks. However, a disadvantage of deep-learning is the requirement for a large amount of fully-annotated training data, which is costly to generate. Weakly-supervised and self-supervised learning is an active research area, but often the model accuracy is inversely correlated with the amount of annotation information provided. Here we focus on a specific subtype of incomplete annotations, which can be generated programmably from experimental data, thus allowing for more annotation information content without sacrificing the annotation speed. We designed a new model architecture for end-to-end training using such incomplete annotations. We benchmarked our method on a variety of publicly available dataset, covering both fluorescence and bright-field imaging modality. We additionally tested our method on a microscopy dataset generated by us, using machine generated annotations. The results demonstrated that our model trained under weak-supervision can achieve segmentation accuracy competitive to, and in some cases surpassing, state-of-the-art models trained under full supervision. Therefore, our method can be a practical alternative to the established full-supervision methods.

cell biology↗