hic_loess: Perform joint loess normalization on two Hi-C datasets

In dozmorovlab/HiCdiff: HiCdiff: Joint normalization and comparative analysis of multiple Hi-C datasets

Description

Perform joint loess normalization on two Hi-C datasets

Usage

hic_loess(hic.table, degree = 1, span = NA, loess.criterion = "gcv",
  Plot = FALSE, parallel = FALSE, BP_param = bpparam(),
  check.differences = FALSE, diff.thresh = "auto", iterations = 10000)

Arguments

hic.table	hic.table or a list of hic.tables generated from the create.hic.table function. list of hic.tables generated from the create.hic.table function. If you want to perform normalization over multiple chromosomes from each cell line at once utilizing parallel computing enter a list of hic.tables and set parallel = TRUE.
degree	Degree of polynomial to be used for loess. Options are 0, 1, 2. The default setting is degree = 1.
span	User set span for loess. If set to NA, the span will be selected automatically using the setting of loess.criterion. Defaults to NA so that automatic span selection is performed.
loess.criterion	Automatic span selection criterion. Can use either 'gcv' for generalized cross-validation or 'aicc' for Akaike Information Criterion. Span selection uses a slightly modified version of the loess.as() function from the fANCOVA package. Defaults to 'gcv'.
Plot	Logical, should the MD plot showing before/after loess normalization be output? Defaults to FALSE.
parallel	Logical, set to TRUE to utilize the parallel package's parallelized computing. Only works on unix operating systems. Only useful if entering a list of hic.tables. Defauts to FALSE.
BP_param	Parameters for BiocParallel. Defaults to bpparam(), see help for BiocParallel for more information http://bioconductor.org/packages/release/bioc/vignettes/BiocParallel/inst/doc/Introduction_To_BiocParallel.pdf
check.differences	Logical, should difference detection be performed? If TRUE, the same procedure as hic_diff will be performed. If FALSE, only normalization will be performed on the entered data. Defaults to FALSE.
diff.thresh	Fold change threshold desired to call a detected difference clinically significant. Set to "auto" by default to indicate that the difference threshold will be automatically calculated as 2 standard deviations of all the adjusted M values. For no p-value adjustment set diff.thresh = NA. To set your own threshold enter a numeric value i.e. diff.thresh = 1. If set to "auto" or a numeric value, a check will be made as follows: if permutation p-value < 0.05 AND M < diff.thresh (the log2 fold change for the difference between IF1 and IF2) then the p-value will be set to 0.5. Defaults to 'auto'.
iterations	Number of iterations for the permuation test. Will only be used if check.differences set to TRUE. Defaults to 10000.

Details

The function takes in a hic.table or a list of hic.table objects created with the create.hic.loess function. If you wish to perform joint normalization on Hi-C data for multiple chromosomes use a list of hic.tables. The process can be parallelized using the parallel setting. The data is fist transformed into what is termed an MD plot (similar to the MA plot/Bland-Altman plot). M is the log difference log2(x/y) between the two datasets. D is the unit distance in the contact matrix. The MD plot can be visualized with the Plot option. Loess regression is then performed on the MD plot to model any biases between the two Hi-C datasets. An adjusted IF is then calculated for each dataset along with an adjusted M. See methods section of Stansfield & Dozmorov 2017 for more details. Note: if you receive the warning "In simpleLoess(y, x, w, span, degree = degree, parametric = parametric, ... :pseudoinverse used..." it should not effect your results, however it can be avoided by manually setting the span to a larger value using the span option.

Value

An updated hic.table is returned with the additional columns of adj.IF1, adj.IF2 for the respective normalized IFs, an adj.M column for the adjusted M, and mc for the loess correction factor. If check.differences is set to TRUE a column containing the p-values for the significance of the difference between the two datasets will also be returned.

Examples

# Create hic.table object using included Hi-C data in sparse upper
# triangular matrix format
data("HMEC.chr22")
data("NHEK.chr22")
hic.table <- create.hic.table(HMEC.chr22, NHEK.chr22, chr= 'chr22')
# Plug hic.table into hic_loess()
result <- hic_loess(hic.table, Plot = TRUE)
# View result
result

dozmorovlab/HiCdiff documentation built on May 20, 2019, 11:13 a.m.