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

FRESNO, CRISTOBAL; GONZALEZ, GERMÁN; MERINO, GABRIELA; FLESIA, GEORGINA; PODHAJCER, OSVALDO LUIS; LLERA, ANDREA SABINA; FERNÁNDEZ, ELMER ANDRÉS

BIOINFORMATICS (OXFORD, ENGLAND)

OXFORD UNIV PRESS

Lugar: Oxford; Año: 2017

1367-4803

eMotivation: irrespectively The of the PAM50 obtained classifier value. is used Nonetheless, to assign r all patients subtype to the correlations highest correlated are required breast to build cancer the subtype risk ofrecurrence (ROR) score, currently used in therapeutic decisions. Present subtype uncertainty estimationsgene are Results: not label accurate, Here permutations. we seldom present considered a Simulations novel single-subject or results require (n=5,228) a non-parametric population-based showed R uncertainty that approach only 61% estimation for subjects this context. based can on be PAM50?s reliablytight ?Assigned" ?Ambiguous" to the correlations PAM50 subtype, between whereas subtypes. 33% The should NA subjects e be ?Not exclusion Assigned" from (NA), the leaving analysis the improved rest toall survival NA subjects subtype curves showed discrimination similar survival yielding behaviour a higher regardless proportion of of the low and v original high PAM50 ROR values. assignment. Conversely, Wepropose to incorporate our PAM50 uncertainty estimation to support therapeutic i decisions.Contact: Availability cristobalfresno@gmail.com and Implementation: Source & efernandez@bdmg.com.ar code can be found in ?pbcmc" R package e at Bioconductor.Supplementary information: Supplementary data are available at Bioinformatics online.w