Dataset Information

Integrating machine learning to construct aberrant alternative splicing event related classifiers to predict prognosis and immunotherapy response in patients with hepatocellular carcinoma.

ABSTRACT: Introduction: In hepatocellular carcinoma (HCC), alternative splicing (AS) is related to tumor invasion and progression. Methods: We used HCC data from a public database to identify AS subtypes by unsupervised clustering. Through feature analysis of different splicing subtypes and acquisition of the differential alternative splicing events (DASEs) combined with enrichment analysis, the differences in several subtypes were explored, cell function studies have also demonstrated that it plays an important role in HCC. Results: Finally, in keeping with the differences between these subtypes, DASEs identified survival-related AS times, and were used to construct risk proportional regression models. AS was found to be useful for the classification of HCC subtypes, which changed the activity of tumor-related pathways through differential splicing effects, affected the tumor microenvironment, and participated in immune reprogramming. Conclusion: In this study, we described the clinical and molecular characteristics providing a new approach for the personalized treatment of HCC patients.

SUBMITTER: Liu W

PROVIDER: S-EPMC9573973 | biostudies-literature | 2022

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Integrating machine learning to construct aberrant alternative splicing event related classifiers to predict prognosis and immunotherapy response in patients with hepatocellular carcinoma.

Liu Wangrui W Zhao Shuai S Xu Wenhao W Xiang Jianfeng J Li Chuanyu C Li Jun J Ding Han H Zhang Hailiang H Zhang Yichi Y Huang Haineng H Wang Jian J Wang Tao T Zhai Bo B Pan Lei L

Frontiers in pharmacology 20221003

<b>Introduction:</b> In hepatocellular carcinoma (HCC), alternative splicing (AS) is related to tumor invasion and progression. <b>Methods:</b> We used HCC data from a public database to identify AS subtypes by unsupervised clustering. Through feature analysis of different splicing subtypes and acquisition of the differential alternative splicing events (DASEs) combined with enrichment analysis, the differences in several subtypes were explored, cell function studies have also demonstrated that ...[more]

PMID: 36263133

Similar Datasets

Project description:BackgroundAlternative splicing (AS) is important in the regulation of gene expression and aberrant AS is emerging as a major factor in the pathogenesis of human conditions, including cancer. Triple-negative breast cancer (TNBC) is the most challenging subtype of breast cancer with strong invasion, high rate of metastasis, and poor prognosis. Here we report a systematic profiling of aberrant AS in TNBC.MethodsThe percent spliced in (PSI) values for AS events in 151 TNBC patients were obtained from The Cancer Genome Atlas (TCGA) SpliceSeq database. Univariate Cox and stepwise Multivariate Cox regression analyses were conducted to find the best prognostic AS model. Splicing regulatory networks were constructed by prognosis-related spliceosome and aberrant AS events. Additionally, pathway enrichment and gene set enrichment analysis (GSEA) were further employed to reveal the significant pathways for prognosis-related AS genes. Finally, splicing regulatory networks were constructed via Spearman's rank correlation coefficients between prognosis-related AS events and splicing factor expressions.ResultsA total of 1,397 prognosis-associated AS events were identified in TNBC. The majority of the parent genes of prognostic AS events exhibited direct interactions to each other in the STRING gene network. Pathways of focal adhesion (p < 0.001), RNA splicing (p = 0.007), homologous recombination (p = 0.042) and ECM-receptor interaction (p = 0.046) were found to be significantly enriched for prognosis-related AS. Additionally, the area under curve (AUC) of the best AS prognostic predictor model reached 0.949, showing a powerful capability to predict outcomes. The Exon Skip (ES) type of AS events displayed more robust and efficient capacity in predicting performance than any other specific AS events type in terms of prognosis. The ES AS signature might confer a strong oncogenic phenotype in the high-risk group with elevated activities in cell cycle and SUMOylating pathways of tumorigenesis, while programmed cell death and metabolism pathways were found to be enriched in the low-risk group of TNBC. The splicing correlation network also revealed a regulatory mode of prognostic splicing factors (SFs) in TNBC.ConclusionOur analysis of AS events in TNBC could not only contribute to elucidating the tumorigenesis mechanism of AS but also provide clues to uncovering underlying prognostic biomarkers and therapeutic targets for further study.

Project description:Prostate cancer (PCa) stands as a prominent contributor to morbidity and mortality among males on a global scale. Cancer-associated fibroblasts (CAFs) are considered to be closely connected to tumour growth, invasion, and metastasis. We explored the role and characteristics of CAFs in PCa through bioinformatics analysis and built a CAFs-based risk model to predict prognostic treatment and treatment response in PCa patients. First, we downloaded the scRNA-seq data for PCa from the GEO. We extracted bulk RNA-seq data for PCa from the TCGA and GEO and adopted "ComBat" to remove batch effects. Then, we created a Seurat object for the scRNA-seq data using the package "Seurat" in R and identified CAF clusters based on the CAF-related genes (CAFRGs). Based on CAFRGs, a prognostic model was constructed by univariate Cox, LASSO, and multivariate Cox analyses. And the model was validated internally and externally by Kaplan-Meier analysis, respectively. We further performed GO and KEGG analyses of DEGs between risk groups. Besides, we investigated differences in somatic mutations between different risk groups. We explored differences in the immune microenvironment landscape and ICG expression levels in the different groups. Finally, we predicted the response to immunotherapy and the sensitivity of antitumour drugs between the different groups. We screened 4 CAF clusters and identified 463 CAFRGs in PCa scRNA-seq. We constructed a model containing 10 prognostic CAFRGs by univariate Cox, LASSO, and multivariate Cox analysis. Somatic mutation analysis revealed that TTN and TP53 were significantly more mutated in the high-risk group. Finally, we screened 31 chemotherapeutic drugs and targeted therapeutic drugs for PCa. In conclusion, we identified four clusters based on CAFs and constructed a new CAFs-based prognostic signature that could predict PCa patient prognosis and response to immunotherapy and might suggest meaningful clinical options for the treatment of PCa.

Dataset Information

Integrating machine learning to construct aberrant alternative splicing event related classifiers to predict prognosis and immunotherapy response in patients with hepatocellular carcinoma.

Publications

Integrating machine learning to construct aberrant alternative splicing event related classifiers to predict prognosis and immunotherapy response in patients with hepatocellular carcinoma.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets