Related Experiment Video
Updated: Mar 9, 2026
11:22
Using Three-color Single-molecule FRET to Study the Correlation of Protein Interactions
Published on: January 30, 2018
10.7K
A novel non-parametric method for uncertainty evaluation of correlation-based molecular signatures: its application
Cristóbal Fresno1, Germán Alexis González1, Gabriela Alejandra Merino1
1UA AREA CS. AGR. ING. BIO. Y S, CONICET, Universidad Católica de Córdoba, Córdoba 5016, Argentina.
Bioinformatics (Oxford, England)
|January 8, 2017
Summary
A new method for breast cancer subtype uncertainty estimation using PAM50 gene permutations reveals that many patients are not reliably assigned. Excluding these "Not Assigned" cases improves risk stratification for recurrence, aiding therapeutic decisions.
Area of Science:
- Genomics
- Bioinformatics
- Oncology
Background:
- The PAM50 classifier assigns breast cancer subtypes but has limitations in accurately estimating subtype uncertainty.
- Current methods for uncertainty estimation are often inaccurate or require population-based approaches, hindering clinical application.
- Accurate subtype correlation is crucial for calculating the risk of recurrence (ROR) score, a key factor in treatment decisions.
Purpose of the Study:
- To develop and validate a novel single-subject non-parametric method for estimating PAM50 classification uncertainty.
- To assess the impact of this uncertainty estimation on breast cancer subtype classification and risk of recurrence (ROR) scoring.
- To propose the integration of this uncertainty estimation into clinical decision-making for breast cancer patients.
Main Methods:
- A novel single-subject non-parametric uncertainty estimation was developed using PAM50 gene label permutations.
- Simulations were conducted on a dataset of 5228 subjects to evaluate the method's performance.
- The impact of excluding uncertain classifications ('Not Assigned' - NA) on survival analysis and ROR score discrimination was analyzed.
Main Results:
- Only 61% of subjects were reliably assigned to a PAM50 subtype; 33% were classified as 'Not Assigned' (NA).
- Excluding NA subjects significantly improved the discrimination of survival subtype curves, leading to a higher proportion of extreme ROR values.
- NA subjects exhibited similar survival behaviors irrespective of their initial PAM50 assignment, highlighting the unreliability of classification for this group.
Conclusions:
- The proposed PAM50 uncertainty estimation method effectively identifies patients with ambiguous subtype correlations.
- Excluding uncertain cases enhances the accuracy of risk stratification for breast cancer recurrence.
- Incorporating this uncertainty estimation can support more informed and personalized therapeutic decisions in breast cancer management.
