miniABS is an absolute, single-sample subtype classifier that can assign subtypes of patients with only the gene expression profile of the patient without relative expression inference (which can ensure reproducibility of subtype assignments).
miniABS has high accuracy without bias for various gene expression quantification technologies.
miniABS predicts subtypes with only 11 biologically meaningful genes in the breast cancer subtype.
The miniABS uses pairwise gene expression ratios (PGER), which is statistically significantly different between subtypes than the expression of a single gene, allowing subtypes to be well classified.
The miniABS does not lose information when subtyping is classified using actual ratio values rather than the magnitude of the expression of the two genes.
Download the source and install it as shown below.
install.packages("miniABS_0.1.1.tar.gz", repos = NULL, type = "source")
library(miniABS)
The Input should be a log2 transformed expression matrix of 11 genes.
data(marker) # list of 11 genes used in miniABS
MYBL2 SFRP1 CEP55 ESR1 FOXA1 MKI67 MLPH PGR ERBB2 FGFR4 KRT17
data(exprRNAseq) # example RNA-seq expression data (eg. log2(FPKM+1))
ratioDir = "directory"
createMatrix(exprRNAseq, ratioDir) #generate a ratio matrix
classifierMiniABS(ratioDir) # classify using 7 models and report final predicted subtypes
A description of the function can be found by entering :
help(package="miniABS")