A novel non-parametric method for uncertainty evaluation of correlation-based molecular signatures: its application on PAM50 algorithm

FRESNO, CRISTÓBAL; GONZÁLEZ, GERMÁN ALEXIS; MERINO, GABRIELA ALEJANDRA; FLESIA, ANA GEORGINA; PODHAJCER, OSVALDO LUIS; LLERA, ANDREA SABINA; FERNÁNDEZ, ELMER ANDRÉS

BIOINFORMATICS (OXFORD, ENGLAND)

OXFORD UNIV PRESS

Año: 2017

1367-4803

Motivation: The PAM50 classifier is used to assign patients to the highest correlated breast cancersubtype irrespectively of the obtained value. Nonetheless, all subtype correlations are required tobuild the risk of recurrence (ROR) score, currently used in therapeutic decisions. Present subtypeuncertainty estimations are not accurate, seldom considered or require a population-based approachfor this context.Results: Here we present a novel single-subject non-parametric uncertainty estimation based onPAM50?s gene label permutations. Simulations results (n ¼ 5228) showed that only 61% subjectscan be reliably ?Assigned? to the PAM50 subtype, whereas 33% should be ?Not Assigned? (NA),leaving the rest to tight ?Ambiguous? correlations between subtypes. The NA subjects exclusionfrom the analysis improved survival subtype curves discrimination yielding a higher proportion oflow and high ROR values. Conversely, all NA subjects showed similar survival behaviour regardlessof the original PAM50 assignment. We propose to incorporate our PAM50 uncertainty estimationto support therapeutic decisions.Availability and Implementation: Source code can be found in ?pbcmc? R package at Bioconductor.