Predicting prostate cancer recurrence: Introducing PCRPS, an advanced online web server

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Prostate cancer (PCa) is a leading cause of cancer death in men.
• Approximately 30% of PCa patients experience biochemical recurrence (BCR) after initial treatment.
• Accurate prediction of PCa recurrence is critical for effective treatment strategies.

Purpose of the Study:

• To develop a reliable tool for predicting PCa recurrence.
• To identify key genes and biomarkers associated with PCa recurrence.
• To create a user-friendly web server for accessing prediction models.

Main Methods:

• Systematic screening and analysis of transcriptomic datasets from 733 PCa patients.
• Meta-analysis, differential gene expression analysis, and feature selection to identify key genes.
• Evaluation of 22 published signatures and development of the Prostate Cancer Recurrence Prediction Signature (PCRPS) using nine machine learning algorithms.

Main Results:

• The PCRPS, based on 23 influential genes, demonstrated superior predictive capability compared to existing signatures and clinicopathological features.
• Random Forest algorithm achieved the highest AUC of 0.72 for PCa recurrence prediction.
• A freely accessible web server (https://urology1926.shinyapps.io/PCRPS/) was developed for utilizing the PCRPS model and predicting drug sensitivities.

Conclusions:

• The developed PCRPS tool offers an effective and reliable method for predicting PCa recurrence.
• This tool can guide clinical management and facilitate individualized treatment strategies for PCa patients.
• The study provides valuable insights into potential drug sensitivities based on PCRPS scores.