Identification of subtype specific biomarkers of clear cell renal cell carcinoma using random forest and greedy algorithm

Identification of Novel Key Genes and Pathways in Multiple Sclerosis Based on Weighted Gene Coexpression Network Analysis and Long Noncoding RNA-Associated Competing Endogenous RNA Network

Objective

Multiple sclerosis (MS) is an autoimmune disease of the central nervous system characterized by chronic inflammation and demyelination. This study is aimed at identifying crucial genes and molecular pathways involved in MS pathogenesis.

Methods

Raw data in GSE52139 were collected from the Gene Expression Omnibus. The top 50% expression variants were subjected to weighted gene coexpression network analysis (WGCNA), and the key module associated with MS occurrence was identified. A long noncoding RNA- (lncRNA-) associated competing endogenous RNA (ceRNA) network was constructed in the key module. The hub gene candidates were subsequently verified in an individual database.

Results

Of the 18 modules obtained, the cyan module was designated as the key module. The established ceRNA network was composed of seven lncRNAs, 45 mRNAs, and 21 microRNAs (miRNAs), and the FAM13A-AS1 was the lncRNA with the highest centrality. Functional assessments indicated that the genes in the cyan module primarily gathered in ribosome-related functional terms. Interestingly, the targeted mRNAs of the ceRNA network enriched in diverse categories. Moreover, highly expressed CYBRD1, GNG12, and SMAD1, which were identified as hub genes, may be associated with “valine leucine and isoleucine degradation,” “base excision repair,” and “fatty acid metabolism,” respectively, according to the results of single gene-based genomes and gene set enrichment analysis (GSEA).

Conclusions

Combined with the WGCNA and ceRNA network, our findings provide novel insights into the pathogenesis of MS. The hub genes discovered herein might also serve as novel biomarkers that correlate with the development and management of MS.

Identification of Prognostic Metabolism-Related Genes in Clear Cell Renal Cell Carcinoma

Background

Clear cell renal cell carcinoma (ccRCC) is a cancer with abnormal metabolism. The purpose of this study was to investigate the effect of metabolism-related genes on the prognosis of ccRCC patients.

Methods

The data of ccRCC patients were downloaded from the TCGA and the GEO databases and clustered using the nonnegative matrix factorization method. The limma software package was used to analyze differences in gene expression. A random forest model was used to screen for important genes. A novel Riskscore model was established using multivariate regression. The model was evaluated based on the metabolic pathway, immune infiltration, immune checkpoint, and clinical characteristics.

Results

According to metabolism-related genes, kidney clear cell carcinoma (KIRC) datasets downloaded from TCGA were clustered into two groups and showed significant differences in prognosis and immune infiltration. There were 667 differentially expressed genes between the two clusters, of which 408 were screened by univariate analysis. Finally, 12 differentially expressed genes (MDK, SLC1A1, SGCB, C4orf3, MALAT1, PILRB, IGHG1, FZD1, IFITM1, MUC20, KRT80, and SALL1) were filtered out using the random forest model. The model of Riskscore was obtained by multiplying the expression levels of these 12 genes with the corresponding coefficients of the multivariate regression. We found that the Riskscore correlated with the expression of these 12 genes; the high Riskscore matched the low survival rate verified in the verification set. The analysis found that the Riskscore model was associated with most of the metabolic processes, immune infiltration of cells such as plasma cells, immune checkpoints such as PD-1, and clinical characteristics such as M stage.

Conclusion

We established a new Riskscore model for the prognosis of ccRCC based on metabolism. The genes in the model provided several novel targets for the study of ccRCC.

RSCMDA: Prediction of Potential miRNA-Disease Associations Based on a Robust Similarity Constraint Learning Method

With the rapid development of biotechnology and computer technology, increasing studies have shown that the occurrence of many diseases in the human body is closely related to the dysfunction of miRNA, and the relationship between them has become a new research hotspot. Exploring disease-related miRNAs information provides a new perspective for understanding the etiology and pathogenesis of diseases. In this study, we proposed a new method based on similarity constrained learning (RSCMDA) to infer disease-associated miRNAs. Considering the problems of noise and incomplete information in current biological datasets, we designed a new framework RSCMDA, which can learn a new disease similarity network and miRNA similarity network based on the existing biological information, and then update the predicted miRNA-disease associations using robust similarity constraint learning method. Consequently, the AUC scores obtained in the global and local cross-validation of RSCMDA are 0.9465 and 0.8494, respectively, which are superior to the other methods. Besides, the prediction performance of RSCMDA is further confirmed by the case study on lung Neoplasms, because 94% of the top 50 miRNAs predicted by the RSCMDA method are confirmed from the existing biological databases or research results. All the results show that RSCMDA is a reliable and effective framework, which can be used as new technology to explore the relationship between miRNA and disease.