# Identification of key genes with differential correlations in prostate cancer

## TL;DR

This study identifies key genes with altered correlations in prostate cancer, offering potential biomarkers for diagnosis and prognosis.

## Contribution

A novel approach combining WGCNA and differential correlation analysis to screen prostate cancer biomarkers.

## Key findings

- 20 gene modules were identified, with three significantly associated with prostate cancer.
- 21 genes were selected as potential biomarkers after intersecting module genes with differentially expressed genes.
- Differential correlations in gene pairs were found during normal-to-tumor transformation.

## Abstract

Background: Prostate cancer, a major global health issue for men, remains a critical clinical challenge in treatment, highlighting the need for improved biomarkers. Treatment options for prostate cancer include active surveillance, surgery, endocrine therapy, chemotherapy, radiotherapy, immunotherapy, etc. However, as the tumor progresses, the effectiveness of treatment regimens gradually decreases. Therefore, we need to understand the biological mechanisms that promote prostate cancer tumorigenesis and progression and to screen biomarkers for diagnosis and prediction of prognosis.

Methods: We utilized the expression profiles of prostate cancer from The Cancer Genome Atlas (TCGA) database and employed weighted gene co-expression network analysis (WGCNA) to construct a gene interaction network. Gene co-expression networks were constructed using WGCNA (soft-threshold power β = 10, scale-free R² > 0.9), with differential correlations computed via Fisher’s z-test (FDR < 0.05). We used the “DiffCorr” package to discriminate between tumor and adjacent normal tissues to identify genes with differential representation in tumor and normal tissues, and perform in-depth analysis of these genes.

Results: Through WGCNA analysis, we identified a total of 20 modules, three gene modules were significantly associated with prostate cancer. We then analyzed the genes in these modules separately by the “DiffCorr” package and intersected these with differentially expressed genes. Finally, 21 genes were screened as biomarkers for prostate cancer.

Conclusions: Our study unveils a prostate cancer tumorigenesis mechanism by identifying differentially correlated gene pairs during normal-to-tumor transformation. We believe that the biomarkers derived from this algorithm have important reference implications for future research in prostate cancer.