make_pairwise_predictions: Predict subtypes, pairwise and by index

In sbuechler/BCCSclassifier: Breast Cancer Consensus Subtype Classifier

Description

This is the most basic step in the process of predicting a subtype for a sample. The BCCSclassifier multi-kTSP object is a list of kTSP models, each consisting of a list of kTSP tests for the best subtype between a pair of possible choices, ranging over all kTSP models in the family. This function applies ktspair prediction to each of these inner models and arranges the results in a tibble, respecting the kTSP model index.

Usage

make_pairwise_predictions(dat, model)

Arguments

dat

An object of class matrix with samples as column names and gene symbols as row names, and entries the corresponding expression levels. The included tibble bccs_classifier_genes contains the gene symbols and ENTREZIDs used in the BCCSclassifier models, along with Affymetrix (hgu133a) and Illumina (illuminaHuman v3) probes that can be used to represent the associated genes. Expression values must not be scaled or median-centered.

model

A character string that is one of "erposneg", "erpos", "erneg", to determine whether BCCSclassifier(ER+/-), BCCSclassifier(ER+), or BCCSclassifier(ER-), respectively, should be used for subtyping.

Details

A function that tests, for each sample, each kTSP model in the family, and each possible pair of subtypes, which subtype is the best choice for the sample

Value

The output is a tibble (data.frame),with columns sample, pair, pair_prediction, index. pair indicates the pair of subtypes tested in the kTSP inner model. pair_prediction is the subtype in the pair chosen by the kTSP model. index (an integer) is the index of the kTSP model in the Multi-kTSP family. This object is called a PPI object, short for pairwise prediction - indexed

sbuechler/BCCSclassifier documentation built on Dec. 22, 2021, 10:19 p.m.