Risk score based on three mRNA expression predicts the survival of bladder cancer

COL10A1 as a Prognostic Biomarker in Association with Immune Infiltration in Prostate Cancer

BACKGROUND

The collagen type X alpha 1 (COL10A1) has recently been found to play an important role in the development and progression of cancer. However, the link between COL10A1 and the tumor immune microenvironment remains understood scantily.

METHODS

In the current study, the pan-cancer data of The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) were used to investigate the expression mode, the clinical prognostic and diagnostic value of COL10A1 in different tumors. We used TCGA data to assess the correlations between COL10A1 and clinical symptoms of prostate cancer. The R packages "edgR" and "clusterProfiler" were used for differential expression gene and enrichment analysis of COL10A1. Immunohistochemistry was further employed to corroborate the expression of COL10A1 gene in prostate cancer. After that, we used TIMER to evaluate the pertinence of COL10A1 expression to immune infiltration level in prostate cancer.

RESULTS

On the whole, COL10A1 was expressed at significantly higher levels in a variety of tumor tissues than in the corresponding normal tissues. Besides, significant correlations with tumor prognosis and relative exactitude in predicting tumors show that COL10A1 may be a probable prognostic and diagnostic biomarker of prostate cancer. In addition, the evidence indicates a significant correlation between COL10A1 and clinical symptoms of prostate cancer. Furthermore, the main molecular functions of COL10A1 included humoral immune response, complement activation, immunoglobulin, regulation of complement activation, and regulation of humoral immune response. Finally, we found that COL10A1 expression is positively correlated with enhanced macrophage and M2 macrophage infiltration in prostate cancer.

CONCLUSION

The study indicates that COL10A1 might participate in M2 macrophage polarization in prostate cancer. COL10A1 might be an innovative biomarker to evaluate tumor microenvironment immune cell infiltration and prognosis in prostate cancer.

Spatial and clonality-resolved 3D cancer genome alterations reveal enhancer-hijacking as a potential prognostic marker for colorectal cancer

The regulatory effect of non-coding large-scale structural variations (SVs) on proto-oncogene activation remains unclear. This study investigated SV-mediated gene dysregulation by profiling 3D cancer genome maps from 40 patients with colorectal cancer (CRC). We developed a machine learning-based method for spatial characterization of the altered 3D cancer genome. This revealed a frequent establishment of "de novo chromatin contacts" that can span multiple topologically associating domains (TADs) in addition to the canonical TAD fusion/shuffle model. Using this information, we precisely identified super-enhancer (SE)-hijacking and its clonal characteristics. Clonal SE-hijacking genes, such as TOP2B, are recurrently associated with cell-cycle/DNA-processing functions, which can potentially be used as CRC prognostic markers. Oncogene activation and increased drug resistance due to SE-hijacking were validated by reconstructing the patient's SV using CRISPR-Cas9. Collectively, the spatial and clonality-resolved analysis of the 3D cancer genome reveals regulatory principles of large-scale SVs in oncogene activation and their clinical implications.

Identification and validation of a five-gene prognostic signature based on bioinformatics analyses in breast cancer

Background

This study aimed to identify prognostic signatures to predict the prognosis of breast cancer (BRCA) patients based on a series of comprehensive analyses of gene expression data.

Methods

The RNA-sequencing expression data and corresponding BRCA patient clinical data were collected from the Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) datasets. Firstly, the differently expressed genes (DEGs) related to prognosis between tumor tissues and normal tissues were ascertained by performing R package "limma". Secondly, the DEGs were used to construct a polygenic risk scoring model by the weighted gene co-expression network analysis (WGCNA) and the least absolute shrinkage and selection operator Cox regression (Lasso-cox) analysis method. Thirdly, survival analysis was performed to investigate the risk score values in the TCGA cohort. And the enrichment analysis, immune cell infiltration levels analysis, and protein-protein internet (PPI) analysis were performed. Simultaneously, the GEO cohort was used to validate the model. Lastly, we constructed a nomogram to explore the influence of polygenic risk score and other clinical factors on the survival probability of patients with BRCA.

Results

A total of 1000 DEGs including 396 upregulated genes and 604 downregulated genes were identified from the TCGA-BRCA dataset. We obtained 5 prognosis-related genes, as the key biomarkers by Lasso-cox analysis (FBXL19, HAGHL, PHKG2, PKMYT1, and TXNDC17), all of which were significantly upregulated in breast tumors. The prognostic prediction of the 5 genes model was great in training and validation cohorts. Moreover, the high-risk group had a poorer prognosis. The Cox regression analysis showed that the comprehensive risk score for 5 genes was an independent prognosis factor.

Conclusion

The 5 genes risk model constructed in this study had an independent predictive ability to distinguish patients with a high risk of death from those with a low-risk score, and it can be used as a practical and reliable prognostic tool for BRCA.

Identification of a Seven-Differentially Expressed Gene-Based Recurrence-Free Survival Model for Melanoma Patients

Melanoma is a malignant tumor that originates in melanocytes of the skin or mucous membrane, which has a high mortality rate and worse prognosis. Therefore, perspective prognosis evaluation seems more important for patients' treatment. Gene expression profiles of melanoma were downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, respectively. 130 consistent differentially expressed genes (DEGs) were identified between melanoma and nevus tissues from two GEO cohorts. Prognostic genes were identified by univariate analysis, and 20 of them were regarded to be associated with the recurrence-free survival (RFS) of melanoma patients. Then, the LASSO Cox regression analysis chose seven of them to establish a seven-DEG-based RFS predicting signature. We demonstrated that this model was more powerful to predict RFS risk than other individual clinical features and was able to independently predict the RFS outcomes in different subsets of patients. We attempted to search for the underlying mechanisms by analyzing the coexpression genes of the seven candidates, and the pathway enrichment analyses indicated that immune response-related pathways might play a critical role in melanoma progression. Finally, we establish a robust seven-DEG-based RFS predicting signature, which will facilitate the personalized treatment of melanoma patients.

Role of FRG1 in predicting the overall survivability in cancers using multivariate based optimal model

FRG1 has a role in tumorigenesis and angiogenesis. Our preliminary analysis showed that FRG1 mRNA expression is associated with overall survival (OS) in certain cancers, but the effect varies. In cervix and gastric cancers, we found a clear difference in the OS between the low and high FRG1 mRNA expression groups, but the difference was not prominent in breast, lung, and liver cancers. We hypothesized that FRG1 expression level could affect the functionality of the correlated genes or vice versa, which might mask the effect of a single gene on the OS analysis in cancer patients. We used the multivariate Cox regression, risk score, and Kaplan Meier analyses to determine OS in a multigene model. STRING, Cytoscape, HIPPIE, Gene Ontology, and DAVID (KEGG) were used to deduce FRG1 associated pathways. In breast, lung, and liver cancers, we found a distinct difference in the OS between the low and high FRG1 mRNA expression groups in the multigene model, suggesting an independent role of FRG1 in survival. Risk scores were calculated based upon regression coefficients in the multigene model. Low and high-risk score groups showed a significant difference in the FRG1 mRNA expression level and OS. HPF1, RPL34, and EXOSC9 were the most common genes present in FRG1 associated pathways across the cancer types. Validation of the effect of FRG1 mRNA expression level on these genes by qRT-PCR supports that FRG1 might be an upstream regulator of their expression. These genes may have multiple regulators, which also affect their expression, leading to the masking effect in the survival analysis. In conclusion, our study highlights the role of FRG1 in the survivability of cancer patients in tissue-specific manner and the use of multigene models in prognosis.

An eight-mRNA signature predicts the prognosis of patients with bladder urothelial carcinoma

Background

Bladder cancer is one of the most common cancers, and its histopathological type is mainly bladder urothelial carcinoma, accounting for about 90%. The prognostic biomarkers of bladder cancer are classified into clinical features biomarkers and molecular biomarkers. Nevertheless, due to the existence of individual specificity, patients with similar pathological characteristics still have great differences in the risk of disease recurrence. Therefore, it is often inaccurate to predict the survival status of patients based on clinical characteristic biomarkers, and a prognostic molecular biomarker that can grade the risk of bladder cancer patients is needed.

Methods

A total of three bladder urothelial carcinoma datasets were used in this study from the Cancer Genome Atlas database and Gene Expression Omnibus database. In order to avoid overfitting, all samples were randomly divided into one training set and three validation sets, which were used to construct and test the prognostic biomarker model of bladder urothelial carcinoma. Univariate and multivariate Cox regression were used to screen candidate mRNAs and construct prognostic biomarkers model. Kaplan-Meier survival analysis and the receiver operating characteristic (ROC) curve were used to evaluate the predictive performance of the model.

Results

A prognostic biomarker model of bladder urothelial carcinoma combining with eight mRNA was constructed. Kaplan-Meier analyses indicated that a significant difference in the survival time of patients between the high-risk and the low-risk group. The area under the ROC curve were 0.632 (95% confidence interval (CI) [0.541-0.723]), 0.693 (95% CI [0.601-0.784]) and 0.686 (95% CI [0.540-0.831]) when the model was used to predict the patient's survival time in three validation datasets. The model showed high accuracy and applicability.

A radiogenomics signature for predicting the clinical outcome of bladder urothelial carcinoma

OBJECTIVES

To determine the integrative value of contrast-enhanced computed tomography (CECT), transcriptomics data and clinicopathological data for predicting the survival of bladder urothelial carcinoma (BLCA) patients.

METHODS

RNA sequencing data, radiomics features and clinical parameters of 62 BLCA patients were included in the study. Then, prognostic signatures based on radiomics features and gene expression profile were constructed by using least absolute shrinkage and selection operator (LASSO) Cox analysis. A multi-omics nomogram was developed by integrating radiomics, transcriptomics and clinicopathological data. More importantly, radiomics risk score-related genes were identified via weighted correlation network analysis and submitted to functional enrichment analysis.

RESULTS

The radiomics and transcriptomics signatures significantly stratified BLCA patients into high- and low-risk groups in terms of the progression-free interval (PFI). The two risk models remained independent prognostic factors in multivariate analyses after adjusting for clinical parameters. A nomogram was developed and showed an excellent predictive ability for the PFI in BLCA patients. Functional enrichment analysis suggested that the radiomics signature we developed could reflect the angiogenesis status of BLCA patients.

CONCLUSIONS

The integrative nomogram incorporated CECT radiomics, transcriptomics and clinical features improved the PFI prediction in BLCA patients and is a feasible and practical reference for oncological precision medicine.

KEY POINTS

• Our radiomics and transcriptomics models are proved robust for survival prediction in bladder urothelial carcinoma patients. • A multi-omics nomogram model which integrates radiomics, transcriptomics and clinical features for prediction of progression-free interval in bladder urothelial carcinoma is established. • Molecular functional enrichment analysis is used to reveal the potential molecular function of radiomics signature.

31 gene expression-based signatures serve as indicators of prognosis for patients with glioma

Glioma has one of the highest mortality rates of all cancer types; however, the prognosis cannot be predicted effectively using clinical indicators, due to the biological heterogeneity of the disease. A total of 31 gene expression-based signatures were identified using selected features in The Cancer Genome Atlas cohorts and machine learning methods. The signatures were assayed in the training dataset and were further validated in four completely independent datasets. Association analyses were implemented, and the results indicated that the signature was not significantly associated with age, radiation therapy or primary tumor size. A nomogram for the 1-year overall survival rate of patients with glioma following initial diagnosis was plotted to facilitate the clinical utilization of the signature. Gene Set Enrichment Analysis was performed based on the signature, in order to determine the potential altered pathways. Metabolic pathways were determined to be significantly enriched. In summary, the 31 gene expression-based signatures were effective and robust in predicting the clinical outcome of glioma in 1,016 glioma samples in five independent international cohorts.

A co-expression network for differentially expressed genes in bladder cancer and a risk score model for predicting survival

Background

Urothelial bladder cancer (BLCA) is one of the most common internal malignancies worldwide with poor prognosis. This study aims to explore effective prognostic biomarkers and construct a prognostic risk score model for patients with BLCA.

Methods

Weighted gene co-expression network analysis (WGCNA) was used for identifying the co-expression module related to the pathological stage of BLCA based on the RNA-Seq data retrieved from The Cancer Genome Atlas database. Prognostic biomarkers screened by Cox proportional hazard regression model and random forest were used to construct a risk score model that can predict the prognosis of patients with BLCA. The GSE13507 dataset was used as the independent testing dataset to test the performance of the risk score model in predicting the prognosis of patients with BLCA.

Results

WGCNA identified seven co-expression modules, in which the brown module consisted of 77 genes was most significantly correlated with the pathological stage of BLCA. Cox proportional hazard regression model and random forest identified TPST1 and P3H4 as prognostic biomarkers. Elevated TPST1 and P3H4 expressions were associated with the high pathological stage and worse survival. The risk score model based on the expression level of TPST1 and P3H4 outperformed pathological stage indicators and previously proposed prognostic models.

Conclusion

The gene co-expression network-based study could provide additional insight into the tumorigenesis and progression of BLCA, and our proposed risk score model may aid physicians in the assessment of the prognosis of patients with BLCA.

Transcriptomic model-based lncRNAs and mRNAs serve as independent prognostic indicators in head and neck squamous cell carcinoma

Head and neck squamous cell carcinoma (HNSC) is one of most common types of cancer worldwide, and mRNAs and long non-coding RNAs (lncRNAs) have been identified as prognostic biomarkers in HNSC. In the present study, using gene expression datasets from multiple platforms, survival-associated genes in HNSC were identified. Subsequently, a combination of 17 genes (14 mRNAs and 3 lncRNA) was optimized using random forest variable hunting and a risk score model for HNSC prognosis was developed using a cohort from The Cancer Genome Atlas. Patients with high-risk scores tend to have earlier disease recurrence and lower survival rates, compared with those with low-risk scores. This observation was further validated in three independent datasets (GSE41613, GSE10300 and E-MTAB-302). Association analysis revealed that the risk score is independent of other clinicopathological observations. On the basis of the results depicted in the nomogram, the risk score performs better in 3-year survival rate prediction than other clinical observations. In summary, the lncRNA-mRNA signature-based risk score successfully predicts the survival of HNSC and serves as an indicator of prognosis.

Deep Learning data integration for better risk stratification models of bladder cancer

We propose an unsupervised multi-omics integration pipeline, using deep-learning autoencoder algorithm, to predict the survival subtypes in bladder cancer (BC). We used TCGA dataset comprising mRNA, miRNA and methylation to infer two survival subtypes. We then constructed a supervised classification model to predict the survival subgroups of any new individual sample. Our training data gave two subgroups with significant survival differences (p-value=8e-4), where high-risk survival subgroup was enriched with KRT6/14 overexpression and PI3K-Akt pathways. We tested the robustness of model by randomly splitting the main dataset into multiple training and test folds, which gave overall significant p-values. Then, we successfully inferred the subtypes for a subset of samples kept as test dataset (p-value=0.03). We further applied our pipeline to predict the survival subgroups from another validation dataset with miRNA data (p-value=0.02). Conclusively, present pipeline is an effective approach to infer the survival subtype of a new sample, exemplified by BC.