In aprodi/methyLImp: Missing Value Estimation of DNA Methylation Data

title: "methyLImp" author: "Pietro Di Lena" date: "r Sys.Date()" output: BiocStyle::html_document vignette: > %\VignetteIndexEntry{methyLImp} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8}

BiocStyle::markdown()

library(knitr)
library(methyLImp)

The `methyLImp` Package

Pietro Di Lena

About the package

The methyLImp implements a missing data imputation method based on single imputation linear regression, especially designed for and tested on DNA methylation data [1].

Installation

The package doesn't have any dependencies from other Bioconductor packages.

Installing the latest package from a local copy (assuming it is in the current working directory of your R session):

install.packages('methyLImp_0.9.9.tar.gz', repos=NULL, type='source')

Trying it out

The package contains a subset of a real 450K Illumina array data, GSE64495, which contains beta values of 100 samples for 200 CpGs with no missing values and it can be used to explore the function quickly:

library('methyLImp')
data(gse64495)
# load in methyLImp dataset
summary(gse64495)

Suggested analysis workflow

Load data

The methylation data array of either beta or M values has to transposed before imputation, as variables need to be on the columns and samples on the rows.

Example workflow

## Load the methyLImp dataset, containing no missing value
data(gse64495)
summary(gse64495)

## Artificially introduce 10% missing values in the first sample
## with the gen_randNA function
set.seed(50)
samp <- 1
frac <- 0.1
gse64495.mis <- gen_randNA(gse64495,samp,frac)
summary(gse64495.mis)

## Impute the missing values with the methyLImp routine.
## Note that variables need to be on the columns and
## samples on the rows.
gse64495.imp <- methyLImp(t(gse64495.mis),min=0,max=0)
gse64495.imp <- t(gse64495.imp)

## Compare imputed and original values
miss <- is.na(gse64495.mis[,samp])
orig <- gse64495[miss,samp]
pred <- gse64495.imp[miss,samp]
gen_stat(orig,pred)