bisect_semi_suprevised: Add together two numbers.

In bisect: Estimating Cell Type Composition from Methylation Sequencing Data

Description

Add together two numbers.

Usage

bisect_semi_suprevised(methylation_unkown_samples, total_reads_unknown_samples,
  methylation_known_samples, total_reads_known_samples,
  cell_composition_known_samples, alpha = NA, iterations = 200)

Arguments

methylation_unkown_samples	a matrix of individuals (rows) on sites (columns), containing the number of methylated reads for each site, in each individual for the samples with unknown cell composition.
total_reads_unknown_samples	a matrix of individuals (rows) on sites (columns), containing the total number of reads for each site, in each individual for the samples with unknown cell composition.
methylation_known_samples	a matrix of individuals (rows) on sites (columns), containing the number of methylated reads for each site, in each individual for the samples with known cell composition.
total_reads_known_samples	a matrix of individuals (rows) on sites (columns), containing the total number of reads for each site, in each individual for the samples with known cell composition.
cell_composition_known_samples	a matrix of individuals (rows) on cell types (columns), containing the proportion of each cell type, in each known sample.
alpha	a vector containing the hyper-parameters for the dirichelt prior. One value for each cell type. If NA, it is initiallized to 1/(number of cell types).
iterations	the number of iterations to use in the EM algorithm.

Value

A list containing P, a matrix of estimated cell proportions for the unknown samples, and Pi, an estimated reference (the probability of methylation in each cell type).

Examples

## Randomly choose samples to be used as known
n_known_samples <- 50
known_samples_indices <- sample.int(nrow(baseline_GSE40279), size = n_known_samples)
known_samples <- as.matrix(baseline_GSE40279[known_samples_indices, ])

## Fit a dirichlet distribution to known samples to use as prior
fit_dirichlet <- sirt::dirichlet.mle(as.matrix(known_samples))
alpha <- fit_dirichlet$alpha

## Prepare the 4 needed matrices
methylation_known <- methylation_GSE40279[known_samples_indices, ]
methylation_unknown <-methylation_GSE40279[-known_samples_indices, ]
total_known <- total_reads_GSE40279[known_samples_indices, ]
total_unknown <- total_reads_GSE40279[-known_samples_indices, ]

## Run Bisect. You should use around 200 iterations. I choose than to accelarate the example.
results <- bisect_semi_suprevised(methylation_unknown, total_unknown,
                                  methylation_known, total_known,
                                  known_samples, alpha, iterations = 10)

bisect documentation built on May 2, 2019, 9:20 a.m.