hepidish {EpiDISH}R Documentation

Hierarchical EpiDISH (HEpiDISH)

Description

HEpiDISH is a iterative hierarchical procedure of EpiDISH. HEpiDISH uses two distinct DNAm references, a primary reference for the estimation of several cell-types fractions, and a separate secondary non-overlapping DNAm reference for the estimation of underlying subtype fractions of one of the cell-type in the primary reference.

Usage

hepidish(beta.m, ref1.m, ref2.m, h.CT.idx, method = c("RPC", "CBS",
  "CP"), maxit = 50, nu.v = c(0.25, 0.5, 0.75),
  constraint = c("inequality", "equality"))

Arguments

beta.m

A data matrix with rows labeling the molecular features (should use same ID as in cent.m) and columns labeling samples (e.g. primary tumour specimens). No missing values are allowed and all values should be positive or zero. In the case of DNA methylation, these are beta-values.

ref1.m

A matrix of primaryreference 'centroids', i.e. representative molecular profiles, for a number of cell subtypes. rows label molecular features (e.g. CpGs,...) and columns label the cell-type. IDs need to be provided as rownames and colnames, respectively. No missing values are allowed, and all values in this matrix should be positive or zero. For DNAm data, values should be beta-values.

ref2.m

Similar to ref1.m, but now a A matrix of secondaryreference. For example, ref1.m contains reference centroids for epithelial cells, fibroblasts and total immune cells. ref2.m can be subtypes of immune cells, such as B-cells, NK cells, monocytes and etc.

h.CT.idx

A index tells which cell-type in ref1.m is the higher order cell-types in ref2.m. For example, ref1.m contains reference centroids for epithelial cells, fibroblasts and total immune cells. ref2.m contains subtypes of immune cells, the h.CT.idx should be 3, corresponding to immune cells in ref1.m.

method

Chioce of a reference-based method ('RPC','CBS','CP')

maxit

Used in RPC mode, the limit on the number of IWLS iterations

nu.v

This is only used for CBS mode. It is a vector of several nv values. nu is parameter needed for nu-classification, nu-regression, and one-classification in svm

constraint

For CP mode, you can choose either of 'inequality' or 'equality' normalization constraint. The default is 'inequality' (i.e sum of weights adds to a number less or equal than 1), which was implemented in Houseman et al (2012).

Value

the estimated cell fraction matrix

References

Zheng SC, Webster AP, Dong D, Feber A, Graham DG, Sullivan R, Jevons S, Lovat LB, Beck S, Widschwendter M, Teschendorff AE A novel cell-type deconvolution algorithm reveals substantial contamination by immune cells in saliva, buccal and cervix. Epigenomics (2018) 10: 925-940. doi: 10.2217/epi-2018-0037.

Teschendorff AE, Breeze CE, Zheng SC, Beck S. A comparison of reference-based algorithms for correcting cell-type heterogeneity in Epigenome-Wide Association Studies. BMC Bioinformatics (2017) 18: 105. doi: 10.1186/s12859-017-1511-5.

Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, Nelson HH, Wiencke JK, Kelsey KT. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics (2012) 13: 86. doi:10.1186/1471-2105-13-86.

Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods (2015) 12: 453-457. doi:10.1038/nmeth.3337.

Examples

data(centEpiFibIC.m)
data(centBloodSub.m)
data(DummyBeta.m)
frac.m <- hepidish(beta.m = DummyBeta.m, ref1.m = centEpiFibIC.m, 
ref2.m = centBloodSub.m, h.CT.idx = 3, method = 'RPC')



[Package EpiDISH version 2.0.0 Index]