Background Forkhead box protein 3 (FOXP3) contributes to tumorigenesis in non-small cell lung cancer (NSCLC), but the biological functions of its isoforms remain unclear. This study investigated the expression patterns and oncogenic roles of two major FOXP3 isoforms (FOXP3Δ3 and FOXP3FL) during NSCLC progression, and clarified the upstream and downstream regulatory mechanisms mediated by FOXP3Δ3. Methods FOXP3Δ3 and FOXP3FL expression was examined in NSCLC tissues and cell lines. Their effects on proliferation, migration, and invasion were assessed using functional assays and xenograft models. Proteomics, the LncTar database, qRT-PCR, and western blot were used to explore the regulatory networks associated with FOXP3Δ3. Results Both FOXP3Δ3 and FOXP3FL were upregulated in NSCLC tissues and cell lines. FOXP3Δ3 played a more dominant role in driving NSCLC malignancy than FOXP3FL. Overexpression of FOXP3Δ3 activated key Hedgehog pathway components (SHH, PTCH1, GLI1). Knockdown of the lncRNA DUXAP8 suppressed NSCLC malignant phenotypes and abrogated FOXP3Δ3-mediated Hedgehog activation. FOXP3Δ3 and DUXAP8 expression correlated significantly with tumor progression and local invasion in NSCLC patients. Conclusions FOXP3Δ3 acts as the predominant isoform that mediates the oncogenic function of FOXP3 in NSCLC. The tumor-promoting mechanism of FOXP3Δ3 is modulated by its upstream regulator DUXAP8 and the subsequent activation of the downstream Hedgehog signaling pathway. However, this study still has limitations. Our clinical cohort was recruited from a single center with a relatively modest sample size, and multicenter cohort studies are warranted to validate the clinical applicability of FOXP3Δ3. In addition, direct molecular evidence supporting the interaction between DUXAP8 and FOXP3Δ3 remains insufficient and requires further functional validation. This work was supported by the National Natural Science Foundation of China (Grant No. 82073054). Database registration number: Not applicable.
Background Per-cell gene-set scoring converts single-cell transcriptomes into interpretable programmes, yet most methods treat gene-set membership as uniform and disregard network context. GSNetAct computes robustly standardised, network-weighted activity scores without modifying the expression matrix. Methods A tuned diffusion weighting (hub damping α = 1.0, two propagation steps, edges restricted to genes detected in ≥ 20% of cells) was assessed by nested donor validation, a topology-specificity null, a topology-versus-selection ablation, parameter sweeps in five cohorts and comparison with eight methods, in interferon-stimulated, tumour-associated, multimodal peripheral blood mononuclear cell (PBMC) and COVID-19 data. Results In an interferon-β stimulation PBMC cohort, tuned diffusion improved condition-state recovery over an unweighted mean by 0.229 normalised mutual information (NMI; 95% CI, 0.200–0.255), a gain retained in held-out donors, and, in the tested benchmark, had the highest interferon-state NMI among nine methods (0.548 versus 0.500), although its held-out lead over the strongest competitor was library-dependent. STRING topology mattered mainly for compact collections, gene selection for larger libraries. Tuned weights were decoupled from within-set degree and concentrated on broadly detected, condition-responsive genes. In an eight-donor identity reference, tuned diffusion underperformed the unweighted mean in adjusted Rand index (ΔARI = −0.018) and degree weighting (-0.020). Conclusions Context-restricted network diffusion improved recovery of a coordinated interferon-stimulated state but is not a universal default; weighting should be chosen per task, retaining unweighted or degree-based aggregation for identity analyses. Limitations: the evidence rests mainly on one interferon-β cohort; parameters were selected on two cohorts; tuned weighting reduced many sets to few genes; most contrasts detected only by tuned diffusion did not differ formally from the unweighted mean; and the COVID-19 analysis was underpowered.
Background Hypertrophic Cardiomyopathy (HCM) is the most prevalent monogenic cardiac disease. Currently, there is no effective cure; management is strictly palliative and focused on symptom relief. Most of its known genetics are sarcomeric, but non-sarcomeric genetics remain poorly characterized. This study used transcriptomics, network biology, and target-based drug knowledge-mining to uncover non-sarcomeric pathological pathways and candidate drugs. Methods Human myocardial RNA-seq data were obtained from the NCBI Sequence Read Archive (Bioproject PRJNA961804). Raw reads were mapped to the human reference genome (GRCh38) using Bowtie2, and quantified with featureCounts. Differential expression was calculated using DESeq2 (|log2FC| ≥ 3, p-adj < 0.05). A protein-protein interaction (PPI) network was constructed by integrating STRING, BioGrid, OmniPath, InWeb_IM and analyzed in Cytoscape using MCODE and CytoHubba. Functional enrichment was done with ShinyGO and Enrichr. Drugs were retrieved from DrugBank for use as candidate drugs. Results 86 high-confidence differentially expressed genes (24 up-, 62 down-regulated) were identified. Two dysregulated pathways were found to converge with enrichment: mineral absorption (down-regulation of both metallothioneins MT1A and MT1M) and neutrophil extracellular trap formation (down-regulation of both FPR1 and FPR2). Hub genes included FPR1, FPR2, CXCR1, CCR1, SSTR5, EDN2, EDN3, S100A8, S100A9, and TLR2. Zinc-containing compounds and formyl-peptide-receptor antagonists (rebamipide, nedocromil) were found through target-based drug knowledge mining. Conclusion These results identify zinc homeostasis and innate immunity as potential non-sarcomeric pathways in HCM. Key caveats of this study include the small sample size, lack of protein level validation and predictive nature of the results of drug knowledge mining. The mechanisms proposed are hypothesis in need of experimental validation.
