Integrated multiomics analysis and machine learning refine molecular subtypes and prognosis for muscle-invasive urothelial cancer

Comprehensive multi-omics analysis of breast cancer reveals distinct long-term prognostic subtypes

Breast cancer (BC) is a leading cause of cancer-related death worldwide. The diverse nature and heterogeneous biology of BC pose challenges for survival prediction, as patients with similar diagnoses often respond differently to treatment. Clinically relevant BC intrinsic subtypes have been established through gene expression profiling and are implemented in the clinic. While these intrinsic subtypes show a significant association with clinical outcomes, their long-term survival prediction beyond 5 years often deviates from expected clinical outcomes. This study aimed to identify naturally occurring long-term prognostic subgroups of BC based on an integrated multi-omics analysis. This study incorporates a clinical cohort of 335 untreated BC patients from the Oslo2 study with long-term follow-up (>12 years). Multi-Omics Factor Analysis (MOFA+) was employed to integrate transcriptomic, proteomic, and metabolomic data obtained from the tumor tissues. Our analysis revealed three prominent multi-omics clusters of BC patients with significantly different long-term prognoses (p = 0.005). The multi-omics clusters were validated in two independent large cohorts, METABRIC and TCGA. Importantly, a lack of prognostic association to long-term follow-up above 12 years in the previously established intrinsic subtypes was shown for these cohorts. Through a systems-biology approach, we identified varying enrichment levels of cell-cycle and immune-related pathways among the prognostic clusters. Integrated multi-omics analysis of BC revealed three distinct clusters with unique clinical and biological characteristics. Notably, these multi-omics clusters displayed robust associations with long-term survival, outperforming the established intrinsic subtypes.

Inflammasome-related gene signatures as prognostic biomarkers in osteosarcoma

Osteosarcoma, the primary bone cancer in adolescents and young adults, is notorious for its aggressive growth and metastatic potential. Our study delved into the prognostic impact of inflammasome-related gene signatures in osteosarcoma patients, employing comprehensive genetic profiling to uncover signatures linked with patient outcomes. We identified three patient subgroups through consensus clustering, with one showing worse survival rates correlated with high FGFR3 and RARB expressions. Immune profiling revealed significant immune cell infiltration differences among these subgroups, affecting survival. Utilising advanced machine learning, including StepCox and gradient boosting machine algorithms, we developed a prognostic model with a notable c-index of 0.706, highlighting CD36 and MYD88 as key genes. Higher inflammasome risk scores from our model were associated with poorer survival, corroborated across datasets. In vitro experiments validated CD36 and MYD88's roles in promoting osteosarcoma cell proliferation, invasion and migration, emphasising their therapeutic potential. This research offers new insights into inflammasomes' role in osteosarcoma, introducing novel biomarkers for risk assessment and potential therapeutic targets. Our findings suggest a pathway towards personalised treatment strategies, potentially improving patient outcomes in osteosarcoma.

Develop a radiomics-based machine learning model to predict the stone-free rate post-percutaneous nephrolithotomy

Radiomics and machine learning have been extensively utilized in the realm of urinary stones, particularly in forecasting stone treatment outcomes. The objective of this study was to integrate clinical variables and radiomic features to develop a machine learning model for predicting the stone-free rate (SFR) following percutaneous nephrolithotomy (PCNL). A total of 212 eligible patients who underwent PCNL surgery at the Second Affiliated Hospital of Nanchang University were included in a retrospective analysis. Preoperative clinical variables and non-contrast-enhanced CT images of all patients were collected, and radiomic features were extracted after delineating the stone ROI. Univariate analysis was conducted to identify clinical variables strongly correlated with the stone-free rate after PCNL, and the least absolute shrinkage and selection operator algorithm (lasso regression) was utilized to screen radiomic features. Four supervised machine learning algorithms, including Logistic Regression, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Gradient Boosting Decision Tree (GBDT), were employed. The clinical variables with strong correlation and screened radiomic features were integrated into the four machine learning algorithms to construct a prediction model, and the receiver operating curve was plotted. The area under the receiver operating curve (AUC), the accuracy rate, the specificity, etc., were used to evaluate the predictive performance of the four models. After analyzing postoperative statistics, the stone-free rate following the procedure was found to be 70.3% (n = 149). Among the various clinical variables examined, factors, such as stone number, stone diameter, stone CT value, stone location, and history of stone surgery, were identified as statistically significant in relation to the stone-free rate after PCNL. A total of 121 radiomic features were extracted, and through lasso regression, 7 features most closely associated with the stone-free rate post-PCNL were identified. The predictive accuracy of different models (Logistic Regression, RF, XGBoost, and GBDT) for determining the stone-free rate after PCNL was evaluated, yielding accuracies of 78.1%, 76.6%, 75.0%, and 73.4%, respectively. The corresponding area under the curve AUC (95%CI) were 0.85 (0.83-0.89), 0.81 (0.76-0.85), 0.82 (0.78-0.85), and 0.77 (0.73-0.81), positioning these models among the top performers in logistic regression prediction. In terms of predictive importance scores, the key factors identified by the logistic regression model were number of stone, zone percentage, stone diameter, and surface area. Similarly, the RF model highlighted number of stone, stone CT value, stone diameter, and surface area as the top predictors. Among the four machine learning models, the logistic regression model demonstrated the highest accuracy and discrimination ability in predicting the stone-free rate following PCNL. In comparison to XGBoost and GBDT, RF also exhibited superior accuracy and a certain level of discrimination ability. However, based on the performance of all four models, logistic regression is more likely to aid in clinical decision-making by assisting clinicians in diagnosing PCNL in patients. This enables us to effectively predict the presence of residual stones post-surgery and ultimately select patients who are suitable candidates for PCNL.

Development of gene panel for predicting recurrence in early-stage cervical cancer patients

Cervical cancer (CC) is a common malignancy affecting women worldwide. Our objective was to develop a consensus-based gene panel using multi-omics data that could effectively predict recurrence in early-stage cervical cancer patients. We utilized the "Multi-Omics Consensus Integration Analysis (MOVICS)" package for consensus clustering design to integrate multiple omics datasets and improve the molecular classification landscape of early-stage CC. We identified the "resting and naive" tumor microenvironment (TME) as cancer subtype (CS) 2. Leveraging the feature genes from the CS classifier, we employed machine learning algorithms to identify a gene panel, including ALDH1A1, CLDN10, MUC13, and C10orf99, which could generate a consensus machine learning-driven score (CMLS) for each patient. Stratifying patients into high and low CMLS groups resulted in Kaplan-Meier curves demonstrating a significant difference in recurrence rates between the two groups. This difference remained significant even after adjusting for clinical features in multivariate Cox regression analysis, with the risk ratio of CMLS surpassing that of clinical characteristics. Furthermore, the TME exhibited notable differences between the different CMLS groups, suggesting that patients with low CMLS may exhibit a better response to immunotherapy. This study highlights the potential of the CMLS approach in predicting recurrence in early-stage cervical cancer patients and provides a screening model for selecting patients suitable for immunotherapy.

Independent prognostic value of CLDN6 in bladder cancer based on M2 macrophages related signature

M2 macrophages are associated with the prognosis of bladder cancer. CLDN6 has been linked to immune infiltration and is crucial for predicting the prognosis in multi-tumor. The effect of CLDN6 on M2 macrophages in bladder cancer remains elusive. Here, we compared a total of 40 machine learning algorithms, then selected optimal algorithm to develop M2 macrophages-related signature (MMRS) based on the identified M2 macrophages related module. MMRS predicted the prognosis better than other models and associated to immunotherapy response. CLDN6, as an important variable in MMRS, was an independent factor for poor prognosis. We found that CLDN6 was highly expressed and affected immune infiltration, immunotherapy response, and M2 macrophages polarization. Meanwhile, CLDN6 promoted the growth of bladder cancer and enhanced the carcinogenic effect by inducing polarization of M2 macrophages. In total, CLDN6 is an independent risk factor in MMRS to predict the prognosis of bladder cancer.

Comprehensive machine learning-based preoperative blood features predict the prognosis for ovarian cancer

PURPOSE

Significant advancements in improving ovarian cancer (OC) outcomes have been limited over the past decade. To predict prognosis and improve outcomes of OC, we plan to develop and validate a robust prognosis signature based on blood features.

METHODS

We screened age and 33 blood features from 331 OC patients. Using ten machine learning algorithms, 88 combinations were generated, from which one was selected to construct a blood risk score (BRS) according to the highest C-index in the test dataset.

RESULTS

Stepcox (both) and Enet (alpha = 0.7) performed the best in the test dataset with a C-index of 0.711. Meanwhile, the low RBS group possessed observably prolonged survival in this model. Compared to traditional prognostic-related features such as age, stage, grade, and CA125, our combined model had the highest AUC values at 3, 5, and 7 years. According to the results of the model, BRS can provide accurate predictions of OC prognosis. BRS was also capable of identifying various prognostic stratifications in different stages and grades. Importantly, developing the nomogram may improve performance by combining BRS and stage.

CONCLUSION

This study provides a valuable combined machine-learning model that can be used for predicting the individualized prognosis of OC patients.

Identification macrophage signatures in prostate cancer by single-cell sequencing and machine learning

BACKGROUND

The tumor microenvironment (TME) encompasses a variety of cells that influence immune responses and tumor growth, with tumor-associated macrophages (TAM) being a crucial component of the TME. TAM can guide prostate cancer in different directions in response to various external stimuli.

METHODS

First, we downloaded prostate cancer single-cell sequencing data and second-generation sequencing data from multiple public databases. From these data, we identified characteristic genes associated with TAM clusters. We then employed machine learning techniques to select the most accurate TAM gene set and developed a TAM-related risk label for prostate cancer. We analyzed the tumor-relatedness of the TAM-related risk label and different risk groups within the population. Finally, we validated the accuracy of the prognostic label using single-cell sequencing data, qPCR, and WB assays, among other methods.

RESULTS

In this study, the TAM_2 cell cluster has been identified as promoting the progression of prostate cancer, possibly representing M2 macrophages. The 9 TAM feature genes selected through ten machine learning methods and demonstrated their effectiveness in predicting the progression of prostate cancer patients. Additionally, we have linked these TAM feature genes to clinical pathological characteristics, allowing us to construct a nomogram. This nomogram provides clinical practitioners with a quantitative tool for assessing the prognosis of prostate cancer patients.

CONCLUSION

This study has analyzed the potential relationship between TAM and PCa and established a TAM-related prognostic model. It holds promise as a valuable tool for the management and treatment of PCa patients.

Robust machine-learning based prognostic index using cytotoxic T lymphocyte evasion genes highlights potential therapeutic targets in colorectal cancer

BACKGROUND

A minute fraction of patients stands to derive substantial benefits from immunotherapy, primarily attributable to immune evasion. Our objective was to formulate a predictive signature rooted in genes associated with cytotoxic T lymphocyte evasion (CERGs), with the aim of predicting outcomes and discerning immunotherapeutic response in colorectal cancer (CRC).

METHODS

101 machine learning algorithm combinations were applied to calculate the CERGs prognostic index (CERPI) under the cross-validation framework, and patients with CRC were separated into high- and low-CERPI groups. Relationship between immune cell infiltration levels, immune-related scores, malignant phenotypes and CERPI were further analyzed. Various machine learning methods were used to identify key genes related to both patient survival and immunotherapy benefits. Expression of HOXC6, G0S2, and MX2 was evaluated and the effects of HOXC6 and G0S2 on the viability and migration of a CRC cell line were in-vitro verified.

RESULTS

The CERPI demonstrated robust prognostic efficacy in predicting the overall survival of CRC patients, establishing itself as an independent predictor of patient outcomes. The low-CERPI group exhibited elevated levels of immune cell infiltration and lower scores for tumor immune dysfunction and exclusion, indicative of a greater potential benefit from immunotherapy. Moreover, there was a positive correlation between CERPI levels and malignant tumor phenotypes, suggesting that heightened CERPI expression contributes to both the occurrence and progression of tumors. Thirteen key genes were identified, and their expression patterns were scrutinized through the analysis of single-cell datasets. Notably, HOXC6, G0S2, and MX2 exhibited upregulation in both CRC cell lines and tissues. Subsequent knockdown experiments targeting G0S2 and HOXC6 resulted in a significant suppression of CRC cell viability and migration.

CONCLUSION

We developed the CERPI for effectively predicting survival and response to immunotherapy in patients, and these results may provide guidance for CRC diagnosis and precise treatment.