Related Experiment Video
Updated: Mar 12, 2026
Constructing and Visualizing Models using Mime-based Machine-learning Framework
Published on: July 22, 2025
Translating a Prognostic DNA Genomic Classifier into the Clinic: Retrospective Validation in 563 Localized Prostate
Emilie Lalonde1, Rached Alkallas2, Melvin Lee Kiang Chua3
1Informatics and Biocomputing Program, Ontario Institute for Cancer Research, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada.
Background:
Localized prostate cancer is clinically heterogeneous, despite clinical risk groups that represent relative prostate cancer-specific mortality. We previously developed a 100-locus DNA classifier capable of substratifying patients at risk of biochemical relapse within clinical risk groups.
Objective:
The 100-locus genomic classifier was refined to 31 functional loci and tested with standard clinical variables for the ability to predict biochemical recurrence (BCR) and metastasis.
Design, Setting, And Participants:
Four retrospective cohorts of radical prostatectomy specimens from patients with localized disease were pooled, and an additional 102-patient cohort used to measure the 31-locus genomic classifier with the NanoString platform.
Outcome Measurements And Statistical Analysis:
The genomic classifier scores were tested for their ability to predict BCR (n=563) and metastasis (n=154), and compared with clinical risk stratification schemes.
Results And Limitations:
The 31-locus genomic classifier performs similarly to the 100-locus classifier. It identifies patients with elevated BCR rates (hazard ratio=2.73, p<0.001) and patients that eventually develop metastasis (hazard ratio=7.79, p<0.001). Combining the genomic classifier with standard clinical variables outperforms clinical models. Finally, the 31-locus genomic classifier was implemented using a NanoString assay. The study is limited to retrospective cohorts.
Conclusions:
The 100-locus and 31-locus genomic classifiers reliably identify a cohort of men with localized disease who have an elevated risk of failure. The NanoString assay will be useful for selecting patients for treatment deescalation or escalation in prospective clinical trials based on clinico-genomic scores from pretreatment biopsies.
Patient Summary:
It is challenging to determine whether tumors confined to the prostate are aggressive, leading to significant undertreatment and overtreatment. We validated a test based on prostate tumor DNA that improves estimations of relapse risk, and that can help guide treatment planning.
