paper.md

title	tags	date	output	authors	bibliography	aas-doi	aas-journal	affiliations
idDMR: Identify Differentially Methylated Regions for Microarray Data	R bioinformatics differentially methylated region kernel smoothing Illumina	September 17, 2022	pdf_document html_document default df_print paged	name corresponding orcid equal-contrib affiliation Daniel Alhassan true 0000-0003-0944-5818 true 1 name orcid equal-contrib affiliation Gayla Olbricht 0000-0002-1213-2241 true 1 name orcid equal-contrib affiliation Akim Adekpedjou 0000-0001-9584-4297 true 1 name orcid equal-contrib affiliation Ebenezer Agbozo 0000-0002-2413-3815 true 2	paper.bib	10.3847/xxxxx <- update this with the DOI from AAS once you know it.	Astrophysical Journal <- The name of the AAS journal.	name index Department of Mathematics and Statistics, Missouri University of Science and Technology, USA 1 name index Department of Big Data Analytics and Methods of VideoAnalytics, Institute of Radioelectronics and Information Technology, Ural Federal University 2

Summary

Identifying and detecting differentially methylated regions (DMRs) in the human genome has been an area of keen research interest for years now. Many methods and their associated R packages have been proposed to solve problems related to DNA methylation data. A few common existing R packages in this area of research are the bumphunter package [@bumphunter;@minfi], the ChAMP package with implements the Probe Lasso method to DMR identification and the DMRcate package [@DMRcate2015]. \par Our knowledge on the co-methylation structure between CpG sites has increased. [@Sun2019] recently revealed that co-methylation patterns within normal tissues were as short as a few hundred base pairs (bp) of each other. [@Sun2019b] in another article on the analysis of methylation patterns between CpG sites in a breast cancer study revealed that co-methylation region lengths differed significantly for unmethylated and methylated states. Another finding worthy of note is that by [@sun2022]. They concluded that the co-methylation patterns on chromosome X were different from the other chromosomes suggesting that the degree of co-methylation may differ on different chromosomes. Further more, the aforementioned methods and their respective packages require the user to set many hyperparameters which leads to inconsistent results in a field so important to human life. In 2019, [@Mallik2019a] performed an extensive simulation study on aforementioned methods and their analysis revealed the lack of consistency in the optimal hyperparameter settings. Also, their findings showed that the methods lacked power to detect DMRs. Owing to the recent findings and the shortcomings of the existing methods, we developed the idDMR R package to help the researcher implement the DMR detection method that uses an array-adaptive normalized kernel-weighted statistic developed by [Alhassan et al (D. Alhassan, G.R. Olbricht, A. Adekpedjou, manuscript submitted)]. This method uses a normalized kernel-weighted statistic to borrow co-methylation information from nearby CpG sites to weight site-level statistics. See [Alhassan et al (D. Alhassan, G.R. Olbricht, A. Adekpedjou, manuscript submitted)] for more details. The advantages of our method include: (1) it requires setting only one hyperparameter and hence leads to consistent results, (2)it increases the sensitivity in detecting true DMRs that may exist in less dense regions while still picking up true DMRs in denser regions \par

idDMR is currently hosted on GitHub. Currently this package contains one DMR detection method in the function aaDMR(). This function uses the array-adaptive normalized kernel statistic in (\ref{A}). The primary arguments it accepts include a CpGsiteAnnotated object, an adapted object , from the originally created cpgannotated object by [@DMRcate2015]; an agglomerate parameter, g, set by the user that is used as the maximum bp distance in collapsing contiguous sites into regions. We describe the key functions in the package and how to use them in the Usage section below.

The estimator at the heart of our DMR detection procedure is the array-adaptive normalized kernel-weighted statistic $S(x_{i})$ below due to [Alhassan et al (D. Alhassan, G.R. Olbricht, A. Adekpedjou, manuscript submitted)]:

\begin{align} S(x_{i}) = Y_{i} + \sum_{j \neq i}^{n}w_{j}(x_{i})Y_{j} \label{A} \end{align} where $Y_{i}$ is the square of limma's t-statistic [@limma] and $w_{j}(x_{i}) = \dfrac{K\left( \dfrac{x_{j}-x_{i}}{h}\right)}{\displaystyle\sum_{j \neq i}^{n}K\left( \dfrac{x_{j}-x_{i}}{h}\right)}$ is a weight measure defined in [Alhassan et al (D. Alhassan, G.R. Olbricht, A. Adekpedjou, manuscript submitted)].

Key Functions

The two core functions of the idDMR package are:

• cpgsite.annotate which runs CpG site-level test i.e. determine differentially methylated loci.

• aadmr which is responsible for identifying differentially methylated regions.

Usage

To use the idDMR package, start by loading dependencies which will install (where necessary) and load all dependent packages:

library(idDMR)

load_dependencies()

To run CpG site-level test (or determine differentially methylated loci), use the function cpgsite.annotate()

myannotation <- cpgsite.annotate(datatype = "array", mval, what = "M", arraytype = "450K",
                                 analysis.type = "differential", design = design_mat,
                                 coef = 2, fdr = 0.05)

Next, use aadmr() function to identify differentially methylated regions.

aadmr = aaDMR(myannotation, g = 1000,  min.cpgs = 2)

#extract the DMResults and output as a dataframe
aadmr_df <- arrange(data.frame(extractRanges(aadmr, genome = "hg19")), seqnames)

References