Construction of a risk prediction model using m6A RNA methylation regulators in prostate cancer: comprehensive bioinformatic analysis and histological validation

Abstract

Background Epigenetic reprogramming reportedly has a crucial role in prostate cancer (PCa) progression. RNA modification is a hot topic in epigenetics, and N6-methyladenosine (m6A) accounts for approximately 60% of RNA chemical modifications. The aim of this study was to evaluate the m6A modification patterns in PCa patients and construct a risk prediction model using m6A RNA regulators. Materials and methods Analyses were based on the levels of 25 m6A regulators in The Cancer Genome Atlas (TCGA). Differentially expressed gene (DEG) and survival analyses were performed according to TCGA-PRAD clinicopathologic and follow-up information. To detect the influences of m6A regulators and their DEGs, consensus clustering analysis was performed, and tumor mutational burden (TMB) estimation and tumor microenvironment (TME) cell infiltration were assessed. mRNA levels of representative genes were verified using clinical PCa data. Results Diverse expression patterns of m6A regulators between tumor and normal (TN) tissues were detected regarding Gleason score (GS), pathological T stage (pT), TP53 mutation, and survival comparisons, with HNRNPA2B1 and IGFBP3 being intersecting genes. HNRNPA2B1 was upregulated in advanced stages (GS > 7, pT3, HR > 1, and TP53 mutation), as verified using clinical PCa tissue. Three distinct m6A modification patterns were identified through consensus clustering analysis, but no significant difference was found among these groups in recurrence-free survival (RFS) analysis. Six DEGs of m6A clusters (m6Aclusters) were screened through univariate Cox regression analysis. MMAB and PAIAP2 were intersecting genes for the five clinical factors. MMAB, which was upregulated in PCa compared with TN, was verified using clinical PCa samples. Three distinct subgroups were established according to the 6 DEGs. Cluster A involved the most advanced stages and had the poorest RFS. The m6A score (m6Ascore) was calculated based on the 6 genes, and the low m6Ascore group showed poor RFS with a negative association with infiltration for 16 of 23 immune-related cells. Conclusion We screened DEGs of m6Aclusters and identified 6 genes (BAIAP2, TEX264, MMAB, JAGN1, TIMM8AP1, and IMP3), with which we constructed a highly predictive model with prognostic value by dividing TCGA-PRAD into three distinct subgroups and performing m6Ascore analysis. This study helps to elucidate the integral effects of m6A modification patterns on PCa progression.

Histone lysine methylation patterns in prostate cancer microenvironment infiltration: Integrated bioinformatic analysis and histological validation

“…This finding implies that clustering TCGA-PRAD patients based on 91 HLM regulators may not be suitable for involvement in a prognostic risk prediction model. We previously divided TCGA-PRAD patients into three clusters based on 25 N6-methyladenosine (m6A) regulators and found no significant differences in RFS analysis (29), which is discordant with the same analysis in gastric cancer (31). This finding suggests the heterogeneity of epigenetic regulation in diverse tumors.…”

“…Recurrence-free survival (RFS) analyses were performed based on the clinical phenotype data of “days_to_first_biochemical_recurrence” and “days_to_last_follow_up.diagnoses”, which we previously described ( 29 ). The survival curves were generated via the Kaplan–Meier method, and statistical significance was evaluated through log-rank tests.…”

“…The clustering results were graphically displayed as a heatmap of the consensus matrices, consensus cumulative distribution function (CDF) plots, and delta area plots. The number of clusters and their stability were determined as previously described ( 29 – 31 ), including the criteria of relatively high consistency within the cluster, low variation coefficient, and no appreciable increase in the area under the CDF curve. PCA was conducted to confirm the fitness and correctness of the HLM modification patterns using the pcromp function of R software.…”

“…The data category of The Cancer Genome Atlas (TCGA) in transcriptome profiling, simple nucleotide variation, copy number variation (CNV), and clinical phenotype were downloaded from the Genomic Data Commons (GDC) Data Portal (https://portal.gdc.cancer.gov/) and University of California Santa Cruz Xena (UCSC) browser (https://xena. ucsc.edu/) as described in our previous study (29).…”

Identification of Ferroptosis-Related Prognostic Signature and Subtypes Related to the Immune Microenvironment for Breast Cancer Patients Receiving Neoadjuvant Chemotherapy

“…Consensus clustering based on Euclidean distance and Ward’s linkage was performed for hierarchical clustering to identify different subtypes using the “ConsensusClusterPlus” package and repeated the procedures 1,000 times to guarantee the stability of the classification ( 34 ). In consideration of a high consistency of clusters, a low coefficient of variation, and no significant increase in the CDF curve, the optimum cluster number could be determined ( 35 ).…”