The application of circulating, cell-free, methylated Septin9 (mSEPT9) DNA in screening and recurrence monitoring is highly promising. CpG island methylator phenotype (CIMP) is associated with microsatellite instability (MSI). The present study was performed to determine the diagnostic accuracy of mSEPT9 for colorectal cancer (CRC) and to evaluate its utility in CRC screening and recurrence monitoring. For screening and diagnosis of CRC, peripheral mSEPT9 detection and fecal occult blood test (FOBT) were performed in 650 subjects, then the level of CEA, CA19–9 and CA724 was quantified in 173 subjects. Clinicopathological parameters and mismatch repair protein were detected among subjects with CRC. For recurrence monitoring of CRC, the sensitivity of mSEPT9 of 70 subjects was compared with tumor markers and contrast enhanced computed tomography (CECT). Seventy-three percent of CRC patients were mSEPT9-positive at 94.5% specificity, and 17.1% of patients with intestinal polyps and adenoma were mSEPT9-positive at 94.5% specificity, which were higher than FOBT for the screening of CRC. The sensitivity and specificity of mSEPT9 for diagnosis and recurrence monitoring were higher than that of CEA, CA19–9 and CA724. The combined detection of mSEPT9 and CECT enhanced the sensitivity for recurrence monitoring. Pre-therapeutic levels of mSEPT9 were strongly associated with TNM stage, Dukes stages and mismatch repair deficiency (dMMR). mSEPT9 analysis might be popularized as a routine biomarker for CRC screening. The combined detection of mSEPT9 and CECT can play an important role for recurrence monitoring. CIMP was highly associated with the pathological stage of CRC and dMMR.
Background: Epithelial ovarian cancer is the leading cause of death from gynecologic cancer, in which serous ovarian carcinoma (SOC) is the most common histological subtype. Although PARP inhibitors (PARPi) and antiangiogenics have been accepted as maintenance treatment in SOC, response to immunotherapy of SOC patients is limited.Methods: The source of transcriptomic data of SOC was from the Cancer Genome Atlas database and Gene Expression Omnibus. The abundance scores of mesenchymal stem cells (MSC scores) were estimated for each sample by xCell. Weighted correlation network analysis is correlated the significant genes with MSC scores. Based on prognostic risk model construction with Cox regression analysis, patients with SOC were divided into low- and high-risk groups. And distribution of immune cells, immunosuppressors and pro-angiogenic factors in different risk groups was achieved by single-sample gene set enrichment analysis. The risk model of MSC scores was further validated in datasets of immune checkpoint blockade and antiangiogenic therapy. In the experiment, the mRNA expression of prognostic genes related to MSC scores was detected by real-time polymerase chain reaction, while the protein level was evaluated by immunohistochemistry.Results: Three prognostic genes (PER1, AKAP12 and MMP17) were the constituents of risk model. Patients classified as high-risk exhibited worse prognosis, presented with an immunosuppressive phenotype, and demonstrated high micro-vessel density. Additionally, these patients were insensitive to immunotherapy and would achieve a longer overall survival with antiangiogenesis treatment. The validation experiments showed that the mRNA of PER1, AKAP12, and MMP17 was highly expressed in normal ovarian epithelial cells compared to SOC cell lines and there was a positive correlation between protein levels of PER1, AKAP12 and MMP17 and metastasis in human ovarian serous tumors.Conclusion: This prognostic model established on MSC scores can predict prognosis of patients and provide the guidance for patients receiving immunotherapy and molecular targeted therapy. Because the number of prognostic genes was fewer than other signatures of SOC, it will be easily accessible on clinic.
Objectives: To develop and validate a hybrid deep learning–machine learning (DL-ML) fusion model for noninvasive preoperative differentiation of solitary pulmonary mucinous adenocarcinoma (SPMA). Methods: A total of 200 patients with pathologically confirmed lung adenocarcinoma, including 37 SPMA cases, treated at Tianjin Union Medical Center between 2018 and 2025 were retrospectively enrolled. Patients were randomly assigned to a training cohort (n = 140) and a test cohort (n = 60). Clinical characteristics, radiomic features, and deep learning features extracted via ResNet50 were integrated. Least absolute shrinkage and selection operator (LASSO) regression was used for feature selection, and multiple machine learning classifiers were compared. Model performance was assessed using the receiver operating characteristic (ROC) curve, decision curve analysis (DCA), and calibration curve analysis (CCA). Results: LASSO regression identified 10 optimal features, comprising 8 radiomic and 2 deep learning features. The random forest classifier yielded the best performance. The hybrid DL-ML fusion model yielded the highest AUC of 0.982 in the training cohort and 0.878 in the test cohort, significantly outperforming the clinical model (Clinic), radiomic model (Rad), deep transfer learning (DTL) model, deep learning–radiomics (DLR) model. In the test cohort, the hybrid DL-ML fusion model achieved an AUC of 0.878, which was significantly higher than that of the clinical model (0.755; p < 0.05). DCA and CCA confirmed favorable clinical utility and calibration. Conclusions: The hybrid DL-ML fusion model enables accurate, noninvasive preoperative differentiation of SPMA. It outperforms conventional clinical assessment and single-modality imaging models, with promising potential for noninvasive preoperative differential diagnosis.