INTRODUCTION: Late‐life depression symptoms are implicated in dementia. We examined how cumulative burden, duration, trajectory, and variability of depression symptoms are associated with incident dementia. METHODS: A prospective cohort analysis of 23,305 dementia‐free adults from four population studies (English Longitudinal Study of Ageing [ELSA], Health and Retirement Study [HRS], Survey of Health, Ageing and Retirement in Europe [SHARE], China Health and Retirement Longitudinal Study [CHARLS]) with repeated Centre for Epidemiologic Studies of Depression scale (CES‐D)/Euro‐Depression scale (EURO‐D) assessments across three waves and a pooled median follow‐up of ≈10.8 years. Exposures included CumSD/cumulative average depression symptom score (CumADS), high‐symptom exposure duration, visit‐to‐visit variability (Standard deviation [SD], coefficient of variation [CV], variation independent of the mean [VIM]), and time‐course patterns. Associations were analyzed using multivariable‐adjusted Cox regression. RESULTS: Each 1‐unit increase in cumulative score was associated with a 3%–8% higher dementia hazard across cohorts. Highest versus lowest cumulative quartiles showed markedly elevated risk. Sustained high exposure for 4 years conferred ≈2.7–3.9× greater risk. Higher variability and worsening trajectories were also linked to higher incidence. Associations were robust across subgroups. DISCUSSION: Persistent and unstable depression symptoms independently predict higher dementia risk, supporting longitudinal mood monitoring and sustained management. Highlights: Multi‐cohort study of 23,305 adults (ELSA, HRS, SHARE, CHARLS). Cumulative depression burden shows a dose–response with dementia risk. Highest versus lowest quartile: dementia hazard up to 18× (HRS). Sustained high symptoms (4 years) linked to ≈2.7–3.9× greater risk. Visit‐to‐visit variability independently associates with higher dementia risk.

Depression and breast cancer (BC) demonstrate significant clinical comorbidity, yet their shared molecular mechanisms remain unclear, particularly regarding immune pathway regulation. This study systematically analyzed Depression-associated gene expression profiles (Gene Expression Omnibus (GEO) database) and BC transcriptomic data (The Cancer Genome Atlas (TCGA) database), identifying overlapping differentially expressed genes (DEGs). Functional enrichment (Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG)) and protein–protein interaction (PPI) network analyses (STRING/Cytoscape) were employed to elucidate biological processes, followed by least absolute shrinkage and selection operator (LASSO) regression and receiver operating characteristic (ROC) curve validation to prioritize key genes. Immune infiltration patterns were assessed via the xCell algorithm, with Spearman correlation linking genes to immune subsets, and single-gene Gene Set Enrichment Analysis (GSEA) evaluating pathway activity. In total, 93 overlapping genes were identified, with predominant involvement in immune-related pathways being revealed by functional enrichment analysis. BHLHE41, EpCAM, and GSTM2 were prioritized as mechanism-associated genes through integrated LASSO regression and ROC analyses. Significant correlations were observed between these genes and specific immune cell populations. GSEA further linked these genes to immune response pathways, suggesting their regulatory roles. These findings highlight immune dysregulation as a shared mechanism underlying Depression-BC comorbidity, providing a foundation for developing early diagnostic strategies and therapeutic strategies targeting both conditions.

Background: Long COVID, a heterogeneous condition characterized by a range of physical and neuropsychiatric presentations, can be presented with a proportion of COVID‐19‐infected individuals.Methods: Transcriptomic data sets of those within gene expression profiles of COVID‐19, long COVID, and healthy controls were retrieved from the GEO database. Differentially expressed genes (DEGs) falling under COVID‐19 and long COVID were identified with R packages, and contemporaneously conducted module detection was performed with the Modular Pharmacology Platform (http://112.86.129.72:48081/). The integration of both DEGs and differentially expressed module‐genes (DEMGs) regarding long COVID and COVID‐19 was intersected by following Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA).Results: There were 11 and 62 differentially expressed modules, 1837 and 179 DEGs, as well as 103 and 508 DEMGs acquiring identified for both COVID‐19 and long COVID, notably enriched in the immune‐correlated signaling pathways. The immune infiltrating cells of long COVID and COVID‐19 were comparatively and respectively assessed via CIBERSORT, ssGSEA, and xCell algorithms. Subsequently, the screening of hub genes involved employing the SVM‐RFE, RF, XGBoost algorithms, and logistic regression analysis. Among the 67 candidate genes were processed with machine learning algorithms and logistic regression, a subgroup consisting of CEP55, CDCA2, MELK, and DEPDC1B, was at last identified as potential biomarkers for predicting the risk of the progression into long COVID after COVID‐19 infections. The predicting performance of the potential biomarkers was quantified with a ROC value of 0.8762542, which proved the combination of potential biomarkers provided the highest performance.Conclusions: In summary, we identified a subgroup of potential biomarkers for predicting the risk of the progression into long COVID after COVID‐19 infection, which could be partly elucidation of the associated molecular mechanisms for long COVID.

Primary sclerosing cholangitis (PSC) is a chronic inflammatory precursor associated with an increased risk of intrahepatic cholangiocarcinoma (ICC), yet identifying malignant features within the persistent inflammatory background remains challenging. In this study, a background-deviation framework was applied to explore malignant-associated determinants during PSC-associated cholangiocarcinogenesis. Single-cell RNA sequencing data from PSC, ICC tumor tissues, and adjacent non-tumor tissues were integrated, followed by functional enrichment, CellChat analysis, Monocle 2 pseudotime reconstruction, Non-negative Matrix Factorization (NMF), STRING/Cytoscape network analysis, and diagnostic signature construction using LASSO regression and exhaustive best subset selection. Single-cell profiling suggested disease-associated cellular remodeling, including cholangiocyte expansion in ICC samples. Functional and intercellular communication analyses indicated a putative transition from an immune-dominant PSC state toward a hyper-biosynthetic ICC-associated phenotype, accompanied by a possible MIF receptor-usage shift from CXCR4 to CD44. Monocle 2 and NMF further identified candidate malignant-associated trajectories and meta-programs, with MYC/TP63-related regulatory signals emerging as potential contributors. Based on these exploratory findings, best subset selection identified a two-gene transcriptomic candidate signature comprising PMAIP1 and GADD45A, which showed promising discriminative performance in internal cross-validation and an external tumor-versus-adjacent validation cohort. These findings provide a transcriptomic basis for further validation of PSC-associated cholangiocarcinogenesis and potential ICC surveillance markers.