ecc: Cell Proportion Estimation using bigmelon

In bigmelon: Illumina methylation array analysis for large experiments

Description

Estimates the relative proprotion of pure cell types within a sample, identical to estimateCellCounts. Currently, only a reference data-set exists for 450k arrays. As a result, if performed on EPIC data, function will convert gds to 450k array dimensions (this will not be memory efficient).

Usage

    estimateCellCounts.gds(
        gds,
        gdPlatform = c("450k", "EPIC", "27k"),
        mn = NULL,
        un = NULL,
        bn = NULL,
        perc = 1,
        compositeCellType = "Blood",
        probeSelect = "auto",
        cellTypes = c("CD8T","CD4T","NK","Bcell","Mono","Gran"),
        referencePlatform = c("IlluminaHumanMethylation450k",
            "IlluminaHumanMethylationEPIC",
            "IlluminaHumanMethylation27k"),
        returnAll = FALSE,
        meanPlot = FALSE,
        verbose=TRUE,
        ...)

Arguments

gds	An object of class gds.class, which contains (un)normalised methylated and unmethylated intensities
gdPlatform	Which micro-array platform was used to analysed samples
mn	'Name' of gdsn node within gds that contains methylated intensities, if NULL it will default to 'methylated' or 'mnsrank' if dasenrank was used prior
un	'Name' of gdsn node within gds that contains unmethylated intensities, if NULL it will default to 'unmethylated' or 'unsrank' if dasenrank was used prior
bn	'Name' of gdsn node within gds that contains un(normalised) beta intensities. If NULL - function will default to 'betas'.
perc	Percentage of query-samples to use to normalise reference dataset. This should be 1 unless using a very large data-set which will allow for an increase in performance
compositeCellType	Which composite cell type is being deconvoluted. Should be either "Blood", "CordBlood", or "DLPFC"
probeSelect	How should probes be selected to distinguish cell types? Options include "both", which selects an equal number (50) of probes (with F-stat p-value < 1E-8) with the greatest magnitude of effect from the hyper- and hypo-methylated sides, and "any", which selects the 100 probes (with F-stat p-value < 1E-8) with the greatest magnitude of difference regardless of direction of effect. Default input "auto" will use "any" for cord blood and "both" otherwise, in line with previous versions of this function and/or our recommendations. Please see the references for more details.
cellTypes	Which cell types, from the reference object, should be we use for the deconvolution? See details.
referencePlatform	The platform for the reference dataset; if the input rgSet belongs to another platform, it will be converted using convertArray.
returnAll	Should the composition table and the normalized user supplied data be return?
verbose	Should the function be verbose?
meanPlot	Whether to plots the average DNA methylation across the cell-type discrimating probes within the mixed and sorted samples.
...	Other arguments, i.e arguments passed to plots

Details

See estimateCellCounts for more information regarding the exact details. estimateCellCounts.gds differs slightly, as it will impose the quantiles of type I and II probes onto the reference Dataset rather than normalising the two together. This is 1) More memory efficient and 2) Faster - due to not having to normalise out a very small effect the other 60 samples from the reference set will have on the remaining quantiles.

Optionally, a proportion of samples can be used to derive quantiles when there are more than 1000 samples in a dataset, this will further increase performance of the code at a cost of precision.

bigmelon documentation built on Nov. 8, 2020, 7:40 p.m.