Metabolomics, a comprehensive omics approach for profiling small-molecule metabolites within biological systems, has become an indispensable tool for disease diagnosis, prognostic assessment, and prediction of therapeutic response. In lung cancer research, metabolomics shows particular promise, notably in early detection, subtype classification, treatment monitoring, and outcome prediction. Lung cancer remains one of the leading causes of cancer-related mortality worldwide, characterized by high treatment costs and persistently low 5-year survival rates, underscoring the urgent need for improved diagnostic strategies. The Lung Cancer Metabolome Database (LCMD) is a pioneering, freely accessible online resource designed to address these challenges by providing a comprehensive repository of lung cancer-related metabolites identified from mass spectrometry-based studies. Lung cancer remains a leading global cause of cancer related mortality, with a critical need for early diagnostic tools. We propose a novel Dual Branch Semi-supervised deep Learning (DBSL) framework to address metabolomics data challenges limited labeled samples and high missingness using the Lung Cancer Metabolome Database (LCMD). We design a global temporal patterns and local feature interactions (a novel Multi-Features Consistency Attention, MFCA).We design a semi-supervised optimization for leveraging pseudo labeling, stochastic augmentation, and consistency loss to exploit unlabeled data. Evaluated under extreme low label regimes (5%, 10%, 15% labeled data), DBSL achieves state of the art performance on the LCMD dataset. These results demonstrate DBSL’s potential for scalable, cost effective lung cancer diagnostics, directly addressing the urgent clinical need for early intervention in Lung Cancer.

BACKGROUND:Numerous studies have shown that esophageal cancer (ESCA) contains areas of intertumoral hypoxia. It is widely accepted that the association of hypoxia with cancer stemness in the tumor microenvironment of ESCA is of profound clinical significance. However, reliable prognostic signatures based on hypoxia and cancer stemness are still lacking in ESCA. MATERIAL AND METHODS:The t-SNE algorithm was used to estimate the hypoxia status based on the transcriptome profiles of the discovery cohort in the TCGA database. Median values of the stemness index were used to group and identify stemness-associated differentially expressed genes (DEGs). The LASSO method and Cox regression model were combined to screen for prognostic genes and to establish a genetic signature based on hypoxia-stemness. The robustness of the prognostic model was then tested in an external independent validation cohort of the GEO database. RESULTS:A total of 8 genes - FBLN2, IL17RB, CYP2W1, AMTN, FABP1, FOXA2, GAS1, and CTSF - were identified to construct a gene signature for ESCA risk stratification. Overall survival was significantly lower in the high-risk group than in the low-risk group in both the internal discovery set and the external validation set. The risk score was found to be an independent prognostic factor for ESCA patients. In addition, a higher risk score was significantly associated with the sensitivity of ESCA patients to gefitinib, bexarotene, dasatinib, and imatinib. CONCLUSIONS:The hypoxia-stemness-based genetic signature established for the first time in our study could be a promising tool for ESCA cancer risk stratification.

Background/Aim: Despite the emergence of cellular, animal, and clinical-based evidence demonstrating a link between hypoxia-inducible factor-1α (HIF-1α) and malignancy, the comprehensive assessment of HIF-1α in pan-cancer patients remains unclear, particularly regarding HIF-1α expression and its association with immune infiltration and immune checkpoint. The present study aimed to investigate the role of HIF-1α expression in various types of malignancies through bioinformatics analysis. Materials and Methods: We investigated the expression and prognostic value of HIF-1α in pan-cancer based on the TCGA (The Cancer Genome Atlas) dataset. The abundance of immune infiltration was estimated by xCell immune deconvolution methods. We investigated the relationship of HIF-1α expression with immune infiltration and immune checkpoint gene expression, with a focus on gastric adenocarcinoma (STAD) and lung squamous cell carcinoma (LUSC). Results: HIF-1α expression had different effects on the prognosis of various cancers. In contrast to the protective effect of HIF-1α expression in LUSC, high levels of HIF-1α expression played a detrimental role in the survival of STAD patients. There was a significant positive correlation between HIF-1α expression and immune infiltration in STAD patients, including regulatory T-cells (Tregs), T-cell CD4+ Th2, neutrophils, M1 and M2 macrophages. In addition, immune checkpoint molecules showed different HIF-1α-related profiles in various carcinomas. Conclusion: A relatively comprehensive view of the oncogenic role of HIF-1α in various tumors based on a pan-cancer analysis is provided in this study. HIF-1α may be considered a poor prognostic biomarker for STAD and, moreover, it may be involved in regulating tumor immune infiltration.