Bioinformatics Analysis of RNA-seq Data Reveals Genes Related to Cancer Stem Cells in Colorectal Cancerogenesis

A Treg-related riskscore model may improve the prognosis evaluation of colorectal cancer

BACKGROUND

Colorectal cancer (CRC) poses a significant health challenge. This study aims to investigate the prognostic value of a regulatory T cell (Treg)-related gene signature in CRC.

METHODS

We extracted the gene expression and clinical data on CRC from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. The gene module related to Treg was identified by weighted gene co-expression network analysis (WGCNA). The genes in the significant module were filtered by univariate Cox, least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression analysis. A riskscore model was established in terms of the key Treg-related genes. The reliability of this riskscore model was validated using the external GEO dataset. The association of riskscore with clinical features, mutation patterns and signaling pathways was explored.

RESULTS

Genes in the blue module showed the strongest association with Tregs. After a series of filtering cycles, seven Treg-related key genes, GDE1, GSR, HSPB1, AOC2, TBX19, TAMM41 and TIGD6, were selected to construct a riskscore model. This model performed well in evaluating the patients' survival in TCGA cohort, and was further affirmed by the GSE17536 validation cohort. For precise evaluation of the patients' survival, we established a nomogram in light of riskscore and clinical factors. Patients in different risk groups had distinct clinical features, mutation patterns and signaling pathway activities. The expression of five key genes was significantly associated with Treg infiltration in the CRC samples.

CONCLUSION

We established a useful riskscore model in light of seven Treg-related genes. This model may contribute to the prognosis evaluation, direct tailored treatment, and hopefully improve clinical outcomes of the CRC patients.

Bioinformatics-based and multiscale convolutional neural network screening of herbal medicines for improving the prognosis of liver cancer: a novel approach

Background

Liver cancer is one of the major diseases threatening human life and health, and this study aims to explore new methods for treating liver cancer.

Methods

A deep learning model for the efficacy of clinical herbal medicines for liver cancer was constructed based on NDCNN, combined with the natural evolutionary rules of a genetic algorithm to obtain the herbal compound for liver cancer treatment. We obtained differential genes between liver cancer tissues and normal tissues from the analysis of TCGA database, screened the active ingredients and corresponding targets of the herbal compound using the TCMSP database, mapped the intersection to obtain the potential targets of the herbal compound for liver cancer treatment in the Venny platform, constructed a PPI network, and conducted GO analysis and KEGG analysis on the targets of the herbal compound for liver cancer treatment. Finally, the key active ingredients and important targets were molecularly docked.

Results

The accuracy of the NDCNN training set was 0.92, and the accuracy of the test set was 0.84. After combining with the genetic algorithm for 1,000 iterations, a set of Chinese herbal compound prescriptions was finally the output. A total of 86 targets of the herbal compound for liver cancer were obtained, mainly five core targets of IL-6, ESR1, JUN, IL1β, and MMP9. Among them, quercetin, kaempferol, and stigmasterol may be the key active ingredients in hepatocellular carcinoma, and the herbal compound may be participating in an inflammatory response and the immune regulation process by mediating the IL-17 signaling pathway, the TNF signaling pathway, and so on. The anticancer effects of the herbal compound may be mediated by the IL-17 signaling pathway, the TNF signaling pathway, and other signaling pathways involved in inflammatory response and immune regulation. Molecular docking showed that the three core target proteins produced stable binding to the two main active ingredients.

Conclusion

The screening of effective herbal compounds for the clinical treatment of liver cancer based on NDCNN and genetic algorithms is a feasible approach and will provide ideas for the development of herbal medicines for the treatment of liver cancer and other cancers.