pbcmc (version 1.0.0)

subtypes,PAM50-method: Subject subtypes for PAM50 adaptation with permuted results.

Description

PAM50 subtypes are obtained using permuted test results. The idea is to give confidence in PAM50 subtype assessment (Perou et al. 2000 & 2010). In this context, the observed Spearman's $\rho$ correlation is tested against the null distribution obtained for each subtype. Then, only significant correlations are used in according to the following scheme:

Not assigned

all subtype have fdr > pcutoff. Hence, there is evidence that the observed $\rho$ can be obtained by random chance.

Assigned

only one fdr <= pcutoff.="" there="" is="" not="" enough="" evidence="" to="" say="" that="" the="" observed="" $\rho$="" does="" belong="" null="" distribution.<="" dd="">

Ambiguous

more than one have fdr <= pcutoff.="" then,="" one="" of="" the="" following="" alternatives="" holds="" given="" result="" $|\rho(profile,="" class_a)-\rho(profile,="" class_b)|="">corCutoff$.

Assigned

If the statement is TRUE.

Ambiguous

If the statement is FALSE.

Under the above scheme, the physician has an objective measurement to support the patient treatment decision. Both, with the given permuted subtype and by interpreting the p-value or fdr of each subtype null distribution test.

Usage

"subtypes"(object, pCutoff = 0.01, corCutoff = 0.1,   where = c("fdr", "pvalue")[1])

Arguments

object

a MolecularPermutationClassifier subclass object.

pCutoff

numeric with p-value/fdr cutoff used depending on "where" selection. Default: 0.01.

corCutoff

numeric with correlation difference between classes cutoff used, i.e., $|\rho(profile, class_A)-\rho(profile, class_B)|>corCutoff$. Default 0.1

where

character with significant value used. Default value is "fdr".

Value

a PAM50 object with the updated slots:

@permutation

$subtype

data.frame with the following fields

$PAM50

the original PAM50 subtype

$Permuted

factor with the following levels:

• "Not assigned": all subtype have fdr > pcutoff
• "Assigned": only one fdr <= pcutoff="" <="" li="">
• "Ambiguous": more than one fdr <= pcutoff="" <="" li="">

$Classes

a character according to "Permuted" field:

• the unique PAM50 subtype if "Assigned"
• a combination for "Ambiguous" or
• NA if "Not assigned".

$Class

idem as Classes but "Ambiguous" is set to PAM50 calls

$Subtype

Classes but "Ambiguous" is kept as "Ambiguous" string

.

@parameters

$pCutoff, $corCutoff and $where are updated accordingly.

References

1. Perou CM, Sorlie T, Eisen MB, et al., 2000, Molecular portraits of human breast tumors. Nature 406:747-752.
2. Perou CM, Parker JS, Prat A, Ellis MJ, Bernard PB., 2010, Clinical implementation of the intrinsic subtypes of breast cancer, The Lancet Oncology 11(8):718-719.

See Also

PAM50 for a complete example.

Other PAM50: as, classify,PAM50-method, filtrate,PAM50-method, pam50centroids, permutate,PAM50-method, subjectReport,PAM50-method

Examples

##Using pam50centroids package example data, which already had been
##filtrated, classified and permutated. 
data(pam50centroids)
summary(pam50centroids)

##Now, let's change pCutoff and corCutoff without the need to run pemutate 
##again
pam50centroids<-subtypes(pam50centroids, pCutoff=0.01, corCutoff=Inf,  
    where="fdr")    
pam50centroids    
summary(pam50centroids)##Note that only Basal is not Ambiguos