R/dreamlet.R
In GabrielHoffman/dreamlet: Scalable differential expression analysis of single cell transcriptomics datasets with complex study designs
Defines functions
as.dreamletResult
Documented in
as.dreamletResult
# Gabriel Hoffman
# April 1, 2021
#
# dreamlet uses linear mixed models in dream to perform differential expression in single cell data
# local definition so methods in this file have this class
#' @return none
setClass("dreamletProcessedData", contains = "list", slots = c(data = "data.frame", metadata = "data.frame", by = "vector"))
#' Class dreamletResult
#'
#' Class \code{dreamletResult} stores results produced by \code{dreamlet()} to give a standard interface for downstream analysis
#'
#' @name dreamletResult-class
#' @rdname dreamletResult-class
#' @exportClass dreamletResult
#' @return none
setClass("dreamletResult", contains = "list", slots = c(df_details = "data.frame", errors = "list", error.initial = "list"))
#' Show object
#'
#' Show object
#'
#' @param object dreamletResult object
#'
#' @return show data stored in object
#'
#' @rdname show-methods
#' @aliases show,dreamletResult,dreamletResult-method
#' @export
setMethod(
"show", "dreamletResult",
function(object) {
object
cat("class:", class(object), "\n")
# assay
nms <- names(object)
if (is.null(nms)) {
nms <- character(length(assays(object, withDimnames = FALSE)))
}
coolcat("assays(%d): %s\n", nms)
df_count <- lapply(object, function(obj) nrow(obj$coefficients))
df_count <- do.call(rbind, df_count)
if (is.null(df_count)) {
cat("No assays retained\n")
} else {
cat("Genes:\n min:", min(df_count[, 1]), "\n max:", max(df_count[, 1]), "\n")
}
# metadata
nms <- names(details(object))
if (is.null(nms)) {
nms <- character(length(metadata(object, withDimnames = FALSE)))
}
coolcat("details(%d): %s\n", nms)
# show coef names
coolcat("coefNames(%d): %s\n", coefNames(object))
# failure rate
df_details <- details(object)
if (nrow(df_details) > 0) {
failure_frac <- sum(df_details$n_errors) / sum(df_details$n_genes)
if (failure_frac > 0) {
txt <- paste0("\nOf ", format(sum(df_details$n_genes), big.mark = ","), " models fit across all assays, ", format(failure_frac * 100, digits = 3), "% failed\n")
cat(txt)
}
}
}
)
#' Print object
#'
#' Print object
#'
#' @param x dreamletResult object
#' @param ... other arguments
#'
#' @return print data stored in object
#'
#' @importFrom utils head tail
#' @importFrom S4Vectors coolcat
#' @export
#' @rdname print-methods
#' @aliases print,dreamletResult,dreamletResult-method
setMethod(
"print", "dreamletResult",
function(x, ...) {
show(x)
}
)
#' Get coefficient names
#'
#' Get coefficient names
#'
#' @param obj A \code{dreamletResult} object
#'
#' @return array storing names of coefficients
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' @rdname coefNames-methods
#' @export
setGeneric("coefNames", function(obj) {
standardGeneric("coefNames")
})
#' @export
#' @rdname coefNames-methods
#' @aliases coefNames,dreamletResult-method
#' @importFrom stats coef
setMethod(
"coefNames", "dreamletResult",
function(obj) {
unique(c(unlist(lapply(obj, function(x) colnames(coef(x))))))
}
)
#' Get error text
#'
#' Get error text
#'
#' @param obj A \code{dreamletResult} object
#'
#' @return \code{tibble} storing error text
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show errors
#' # but none are reported
#' res.err = seeErrors(res.dl)
#'
#' @rdname seeErrors-methods
#' @export
setGeneric("seeErrors", function(obj) {
standardGeneric("seeErrors")
})
#' @export
#' @rdname seeErrors-methods
#' @aliases seeErrors,dreamletResult-method
#' @importFrom dplyr as_tibble
setMethod(
"seeErrors", "dreamletResult",
function(obj) {
# Initial fit
df <- lapply(names(obj@error.initial), function(id) {
if (length(obj@error.initial[[id]]) == 0) {
return(NULL)
}
tibble(
assay = id,
errorTextInitial = obj@error.initial[[id]]
)
})
df <- bind_rows(df)
txt = paste(" Assay-level errors:", nrow(df))
message(txt)
# Gene-level
df2 <- lapply(names(obj@errors), function(id) {
if (length(obj@errors[[id]]) == 0) {
return(NULL)
}
tibble(
assay = id,
feature = names(obj@errors[[id]]),
errorText = obj@errors[[id]]
)
})
df2 <- bind_rows(df2)
txt = paste(" Gene-level errors:", nrow(df2))
message(txt)
list(assayLevel = df, geneLevel = df2)
}
)
setGeneric("assayNames", SummarizedExperiment::assayNames)
setGeneric("assay", SummarizedExperiment::assay)
# setGeneric('colData', SummarizedExperiment::colData)
# setGeneric('metadata', S4Vectors::metadata)
#' Get assayNames
#'
#' Get assayNames
#'
#' @param x dreamletResult object
#' @param ... other arguments
#'
#' @rdname assayNames-methods
#' @aliases assayNames,dreamletResult,dreamletResult-method
#' @export
setMethod(
"assayNames", signature(x = "dreamletResult"),
function(x, ...) {
names(x)
}
)
#' Get assay
#'
#' Get assay
#'
#' @param x dreamletResult object
#' @param i number indicating index, or string indicating assay
#' @param withDimnames not used
#' @param ... other arguments
#'
#' @rdname assay-methods
#' @aliases assay,dreamletResult,dreamletResult-method
#' @export
setMethod(
"assay", signature(x = "dreamletResult"),
function(x, i, withDimnames = TRUE, ...) {
x[[i]]
}
)
#' Subset with brackets
#'
#' Subset with brackets
#'
#' @param x dreamletResult object
#' @param i indeces to extract
#'
#' @rdname extract-methods
#' @aliases [,dreamletResult,dreamletResult-method
#' @export
setMethod(
"[", signature(x = "dreamletResult"),
function(x, i) {
res <- new("dreamletResult", x@.Data[i],
df_details = details(x)[i, , drop = FALSE]
)
names(res) <- names(x)[i]
res
}
)
#' Convert list of regression fits to \code{dreamletResult}
#'
#' Convert list of regression fits to \code{dreamletResult} for downstream analysis
#'
#' @param fitList list of regression fit with \code{dream()}
#' @param df_details \code{data.frame} storing assay details
#'
#' @details Useful for combining multiple runs of \code{dreamletCompareClusters()} into a single \code{dreamletResult} for downstream analysis
#' @return object of class \code{dreamletResult}
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # first comparison
#' ct.pairs <- c("B cells", "CD14+ Monocytes")
#' fit <- dreamletCompareClusters(pb, ct.pairs, method = "fixed")
#'
#' # second comparison
#' ct.pairs2 <- c("B cells", "CD8 T cells")
#' fit2 <- dreamletCompareClusters(pb, ct.pairs2, method = "fixed")
#'
#' # Make a list storing each result with a meaningful name
#' fitList <- list()
#'
#' id <- paste0("[", ct.pairs[1], "]_vs_[", ct.pairs[2], "]")
#' fitList[[id]] <- fit
#'
#' id <- paste0("[", ct.pairs2[1], "]_vs_[", ct.pairs2[2], "]")
#' fitList[[id]] <- fit2
#'
#' # create a dreamletResult form this list
#' res.compare <- as.dreamletResult(fitList)
#' res.compare
#'
#' @importFrom methods new is
#' @export
as.dreamletResult <- function(fitList, df_details = NULL) {
# check that names are not empty
if (any(names(fitList) == "")) {
stop("names(fitList) must not contain empty names")
}
# check that names are unique and not empty
if (length(unique(names(fitList))) != length(fitList)) {
stop("names(fitList) must be unique")
}
# check that entries are the result of a dream() fit
if (any(!vapply(fitList, is, class2 = "MArrayLM", FUN.VALUE = logical(1)))) {
stop("Each element of fitList must be a fit from dream()")
}
if (is.null(df_details)) {
res <- new("dreamletResult", fitList)
} else {
res <- new("dreamletResult", fitList, df_details = df_details)
}
res
}
#' Table of Top Genes from dreamlet fit
#'
#' Extract a table of the top-ranked genes from a dreamlet fit.
#'
#' @param fit dreamletResult object
#' @param coef coef
#' @param number number
#' @param genelist genelist
#' @param adjust.method adjust.method
#' @param sort.by sort.by
#' @param resort.by resort.by
#' @param p.value p.value
#' @param lfc lfc
#' @param confint confint
#'
#' @return \code{data.frame} storing hypothesis test for each gene and cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' topTable(res.dl, coef = "group_idstim", number = 3)
#'
#' @seealso \code{limma::topTable()}, \code{variancePartition::topTable()}
#' @rdname topTable-methods
#' @aliases topTable,dreamletResult,dreamletResult-method
#' @importFrom gtools smartbind
#' @export
setMethod(
"topTable", signature(fit = "dreamletResult"),
function(fit,
coef = NULL,
number = 10,
genelist = NULL,
adjust.method = "BH",
sort.by = "P",
resort.by = NULL,
p.value = 1,
lfc = 0,
confint = FALSE) {
if (any(!coef %in% coefNames(fit))) {
stop("coef must be in coefNames")
}
# Run topTable on each assay
res <- lapply(assayNames(fit), function(k) {
fit1 <- assay(fit, k)
# if entries in coef are found
if (any(coef %in% colnames(coef(fit1)))) {
# only test coefs estimated in this model
coef_found <- coef[coef %in% colnames(coef(fit1))]
# specifiying genelist create column ID
tab <- topTable(fit1, coef = coef_found, number = Inf, sort.by = "none", genelist = rownames(fit1), confint = confint)
# with multiple coefs, first column is 'ProbeID'
if (colnames(tab)[1] == "ProbeID") {
colnames(tab)[1] <- "ID"
}
if (nrow(tab) > 0) {
tab <- tab[!is.na(tab$ID), , drop = FALSE]
# if doesn't have z.std, add it
if (!"z.std" %in% colnames(tab)) {
tab$z.std <- tab$t
}
res <- data.frame(assay = k, tab)
} else {
res <- NULL
}
} else {
res <- NULL
}
res
})
# combine across assays
# allow columns to be missing when coef is array
# and some cell types dont have all of them
# Update: handle case when coef is not found
i <- which(!vapply(res, is.null, logical(1)))
if (length(i) > 1) {
res <- smartbind(list = res[i])
} else {
res <- res[i]
}
res <- DataFrame(res)
# remove rownames
rownames(res) <- c()
if (nrow(res) == 0) {
stop("No results were found matching given criteria")
}
# apply multiple testing across *all* tests
# subset based on number afterwards
res$adj.P.Val <- p.adjust(res$P.Value, adjust.method)
# apply filtering afterwards
res <- res[res$P.Value <= p.value, , drop = FALSE]
# check that logFC exists,
# since with multiple coefs it does not
if ("logFC" %in% colnames(res) & lfc != 0 & length(coef) == 1) {
res <- res[abs(res$logFC) >= lfc, , drop = FALSE]
}
if (!is.null(genelist)) {
res <- res[res$ID %in% genelist, , drop = FALSE]
}
opt <- c("logFC", "AveExpr", "P", "t", "B", "none")
if (!sort.by %in% opt) {
cmd <- paste("sort.by must be in: ", paste0(opt, collapse = ", "))
stop(cmd)
}
if( all(c("t", "F.std") %in% colnames(res)) ){
warning("Mixture of univariate and multivariate results is returned")
}
# sorting
ord <- switch(sort.by,
logFC = order(abs(res$logFC), decreasing = TRUE),
AveExpr = order(res$AveExpr, decreasing = TRUE),
P = order(res$P.Value, decreasing = FALSE),
t = order(abs(t), decreasing = TRUE),
B = order(res$B, decreasing = TRUE),
none = seq_len(nrow(res))
)
head(res[ord, ], number)
}
)
#' Test if coefficient is different from a specified value
#'
#' Test if coefficient is different from a specified value
#'
#' @param fit dreamletResult object
#' @param lfc a minimum log2-fold-change below which changes not considered scientifically meaningful
#' @param coef which coefficient to test
#' @param number number of genes to return
#' @param sort.by column to sort by
#'
#' @return \code{DataFrame} storing hypothesis test for each gene and cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' getTreat(res.dl, coef = "group_idstim", number = 3)
#'
#' @seealso \code{limma::topTreat()}, \code{variancePartition::getTreat()}
#' @importFrom variancePartition getTreat
#' @rdname getTreat-methods
#' @aliases getTreat,dreamletResult,dreamletResult-method
#' @export
setMethod(
"getTreat", signature(fit = "dreamletResult"),
function(fit, lfc = log2(1.2), coef = NULL, number = 10, sort.by = "p") {
if (any(!coef %in% coefNames(fit))) {
stop("coef must be in coefNames")
}
adjust.method <- "BH"
# Run topTable on each assay
res <- lapply(assayNames(fit), function(k) {
fit1 <- assay(fit, k)
# if coef is not found
if (all(coef %in% colnames(coef(fit1)))) {
tab <- getTreat(fit1,
lfc = lfc,
coef = coef,
number = Inf
)
res <- data.frame(assay = k, tab)
} else {
res <- NULL
}
res
})
# combine across assays
res <- DataFrame(do.call(rbind, res))
# remove rownames
rownames(res) <- c()
# apply multiple testing across *all* tests
# subset based on number afterwards
res$adj.P.Val <- p.adjust(res$P.Value, adjust.method)
opt <- c("logFC", "AveExpr", "P", "t", "B", "none")
if (!tolower(sort.by) %in% tolower(opt)) {
cmd <- paste("sort.by must be in: ", paste0(opt, collapse = ", "))
stop(cmd)
}
# sorting
ord <- switch(tolower(sort.by),
logfc = order(abs(res$logFC), decreasing = TRUE),
aveexpr = order(res$AveExpr, decreasing = TRUE),
p = order(res$P.Value, decreasing = FALSE),
t = order(abs(t), decreasing = TRUE),
b = order(res$B, decreasing = TRUE),
none = seq_len(nrow(res))
)
head(res[ord, ], number)
}
)
#' Differential expression for each assay
#'
#' Perform differential expression for each assay using linear (mixed) models
#'
#' @param x SingleCellExperiment or dreamletProcessedData object
#' @param formula regression formula for differential expression analysis
#' @param data metadata used in regression formula
#' @param assays array of assay names to include in analysis. Defaults to \code{assayNames(x)}
#' @param contrasts character vector specifying contrasts specifying linear combinations of fixed effects to test. This is fed into \code{makeContrastsDream( formula, data, contrasts=contrasts)}
#' @param min.cells minimum number of observed cells for a sample to be included in the analysis
# @param isCounts logical, indicating if data is raw counts
#' @param robust logical, use eBayes method that is robust to outlier genes
# @param normalize.method normalization method to be used by \code{calcNormFactors}
#' @param quiet show messages
#' @param BPPARAM parameters for parallel evaluation
#' @param use.eBayes should \code{eBayes} be used on result? (defualt: TRUE)
#' @param ... other arguments passed to \code{dream}
#'
#' @details
#' Fit linear (mixed) model on each cell type separately. For advanced use of contrasts see \code{variancePartition::makeContrastsDream()} and vignette \url{https://gabrielhoffman.github.io/variancePartition/articles/dream.html#advanced-hypothesis-testing-1}.
#'
#' @return Object of class \code{dreamletResult} storing results for each cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # Examine results
#' res.dl
#'
#' # Examine details for each assay
#' details(res.dl)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' topTable(res.dl, coef = "group_idstim", number = 3)
#'
#' @importFrom BiocParallel SerialParam
#' @importFrom SummarizedExperiment colData assays
#' @importFrom S4Vectors as.data.frame
#' @importFrom Rdpack reprompt
#' @seealso \code{variancePartition::dream()}, \code{variancePartition::makeContrastsDream()}
#' @export
setGeneric(
"dreamlet",
function(x, formula, data = colData(x), assays = assayNames(x), contrasts = NULL, min.cells = 10, robust = FALSE, quiet = FALSE, BPPARAM = SerialParam(), use.eBayes = TRUE, ...) {
standardGeneric("dreamlet")
}
)
#' @importFrom variancePartition getContrast dream makeContrastsDream
#' @importFrom SummarizedExperiment colData assays
#' @importFrom S4Vectors as.data.frame
#' @export
#' @rdname dreamlet
#' @aliases dreamlet,dreamletProcessedData-method
setMethod(
"dreamlet", "dreamletProcessedData",
function(x, formula, data = colData(x), assays = assayNames(x), contrasts = NULL, min.cells = 10, robust = FALSE, quiet = FALSE, BPPARAM = SerialParam(), use.eBayes = TRUE, ...) {
# checks
stopifnot(is(formula, "formula"))
# check if assays are valid
if (any(!assays %in% assayNames(x))) {
idx <- which(!assays %in% assayNames(x))
txt <- paste("Assays are not found in dataset:", paste(head(assays[idx]), collapse = ", "))
stop(txt)
}
# extract metadata shared across assays
data_constant <- as.data.frame(data)
# remove samples with missing covariate data
idx <- lapply(all.vars(formula), function(v) {
which(is.na(data_constant[[v]]))
})
idx <- unique(unlist(idx))
if (length(idx) > 1) {
data_constant <- droplevels(data_constant[-idx, , drop = FALSE])
}
# for each assay
resList <- lapply(assays, function(k) {
if (!quiet) message(" ", k, "...", appendLF = FALSE)
startTime <- Sys.time()
geneExpr <- assay(x, k)
# get names of samples to extract from
# intersecting between geneExpr and metadata
ids <- intersect(colnames(geneExpr), rownames(data_constant))
geneExpr <- geneExpr[, ids, drop = FALSE]
# merge data_constant (data constant for all cell types)
# with metadata(sceObj)$aggr_means (data that varies)
data2 <- merge_metadata(
data_constant[ids, , drop = FALSE],
metadata(x),
k,
x@by
)
# drop any constant terms from the formula
form_mod <- removeConstantTerms(formula, data2)
# Drop variables in a redundant pair
form_mod <- dropRedundantTerms(form_mod, data2)
# initialize with no errors
errorArray <- NULL
# drop any constant terms from the formula
# if there is at least one variable or an intercept
tst1 = (length(all.vars(form_mod)) > 0) |
isTRUE(all.equal(form_mod, ~1))
if (tst1 & isFullRank(form_mod, data2)) {
# get contrasts customized for the formula for this cell type
if (!is.null(contrasts)) {
L <- makeContrastsDream(form_mod, data2, contrasts = contrasts, nullOnError = TRUE)
} else {
L <- NULL
}
fit <- dream(geneExpr, form_mod, data2, L = L, BPPARAM = BPPARAM, ..., hideErrorsInBackend = TRUE)
errorArray <- attr(fit, "errors")
if (!is.null(attr(fit, "error.initial"))) {
if (!quiet) message(format(Sys.time() - startTime, digits = 2))
return(list(
fit = NULL,
formula = form_mod,
n_retain = ncol(geneExpr),
errors = errorArray,
error.initial = attr(fit, "error.initial")
))
}
# if model is degenerate
if (!is.null(fit) && !any(is.na(fit$sigma))) {
if (!is.null(fit$rdf)) {
# keep genes with residual degrees of freedom > 1
# this prevents failures later
keep <- (fit$rdf >= 1)
fit <- fit[keep, ]
if (nrow(fit) == 0) fit <- NULL
}
if (use.eBayes & !is.null(fit)) {
# use counts directly if is an EList
isCounts <- ifelse(is(geneExpr, "EList"), TRUE, FALSE)
# borrow information across genes with the empirical Bayes step
fit <- eBayes(fit, robust = robust, trend = !isCounts)
}
} else {
fit <- NULL
}
} else {
fit <- NA
attr(fit, "error.initial") = "Design matrix is singular, covariates are very correlated"
}
if (!quiet) message(format(Sys.time() - startTime, digits = 2))
err.init = attr(fit, "error.initial")
if( !is(fit, "MArrayLM") && is.na(fit) ) fit <- NULL
list(
fit = fit,
formula = form_mod,
n_retain = ncol(geneExpr),
errors = errorArray,
error.initial = err.init
)
})
# name each result by the assay name
names(resList) <- assays
# extract fit
fitList <- lapply(resList, function(x) x$fit)
# only keep entries that are not NULL
# NUll is returned when coef of interest is dropped
fitList <- fitList[!vapply(fitList, is.null, FUN.VALUE = logical(1))]
# Handle errors
#--------------
# get error messages
error.initial <- lapply(resList, function(x) {
x$error.initial
})
names(error.initial) <- names(resList)
errors <- lapply(resList, function(x) {
x$errors
})
names(errors) <- names(resList)
# extract details
df_details <- lapply(names(resList), function(id) {
data.frame(
assay = id,
n_retain = resList[[id]]$n_retain,
formula = Reduce(paste, deparse(resList[[id]]$formula)),
formDropsTerms = !equalFormulas(resList[[id]]$formula, formula),
n_genes = nrow(x[[id]]), # original genes
# n_genes = nrow(resList[[id]]$fit),
n_errors = length(resList[[id]]$errors),
error_initial = ifelse(is.null(resList[[id]]$error.initial), FALSE, TRUE)
)
})
df_details <- do.call(rbind, df_details)
ndrop <- sum(df_details$formDropsTerms)
if (ndrop > 0) {
warning("Terms dropped from formulas for ", ndrop, " assays.\n Run details() on result for more information")
}
failure_frac <- sum(df_details$n_errors) / sum(df_details$n_genes)
if (failure_frac > 0) {
txt <- paste0("\nOf ", format(sum(df_details$n_genes), big.mark = ","), " models fit across all assays, ", format(failure_frac * 100, digits = 3), "% failed\n")
message(txt)
}
new("dreamletResult", fitList,
df_details = df_details,
errors = errors,
error.initial = error.initial
)
}
)
GabrielHoffman/dreamlet documentation built on July 20, 2024, 9:03 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions as.dreamletResult
Documented in as.dreamletResult
# Gabriel Hoffman
# April 1, 2021
#
# dreamlet uses linear mixed models in dream to perform differential expression in single cell data
# local definition so methods in this file have this class
#' @return none
setClass("dreamletProcessedData", contains = "list", slots = c(data = "data.frame", metadata = "data.frame", by = "vector"))
#' Class dreamletResult
#'
#' Class \code{dreamletResult} stores results produced by \code{dreamlet()} to give a standard interface for downstream analysis
#'
#' @name dreamletResult-class
#' @rdname dreamletResult-class
#' @exportClass dreamletResult
#' @return none
setClass("dreamletResult", contains = "list", slots = c(df_details = "data.frame", errors = "list", error.initial = "list"))
#' Show object
#'
#' Show object
#'
#' @param object dreamletResult object
#'
#' @return show data stored in object
#'
#' @rdname show-methods
#' @aliases show,dreamletResult,dreamletResult-method
#' @export
setMethod(
"show", "dreamletResult",
function(object) {
object
cat("class:", class(object), "\n")
# assay
nms <- names(object)
if (is.null(nms)) {
nms <- character(length(assays(object, withDimnames = FALSE)))
}
coolcat("assays(%d): %s\n", nms)
df_count <- lapply(object, function(obj) nrow(obj$coefficients))
df_count <- do.call(rbind, df_count)
if (is.null(df_count)) {
cat("No assays retained\n")
} else {
cat("Genes:\n min:", min(df_count[, 1]), "\n max:", max(df_count[, 1]), "\n")
}
# metadata
nms <- names(details(object))
if (is.null(nms)) {
nms <- character(length(metadata(object, withDimnames = FALSE)))
}
coolcat("details(%d): %s\n", nms)
# show coef names
coolcat("coefNames(%d): %s\n", coefNames(object))
# failure rate
df_details <- details(object)
if (nrow(df_details) > 0) {
failure_frac <- sum(df_details$n_errors) / sum(df_details$n_genes)
if (failure_frac > 0) {
txt <- paste0("\nOf ", format(sum(df_details$n_genes), big.mark = ","), " models fit across all assays, ", format(failure_frac * 100, digits = 3), "% failed\n")
cat(txt)
}
}
}
)
#' Print object
#'
#' Print object
#'
#' @param x dreamletResult object
#' @param ... other arguments
#'
#' @return print data stored in object
#'
#' @importFrom utils head tail
#' @importFrom S4Vectors coolcat
#' @export
#' @rdname print-methods
#' @aliases print,dreamletResult,dreamletResult-method
setMethod(
"print", "dreamletResult",
function(x, ...) {
show(x)
}
)
#' Get coefficient names
#'
#' Get coefficient names
#'
#' @param obj A \code{dreamletResult} object
#'
#' @return array storing names of coefficients
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' @rdname coefNames-methods
#' @export
setGeneric("coefNames", function(obj) {
standardGeneric("coefNames")
})
#' @export
#' @rdname coefNames-methods
#' @aliases coefNames,dreamletResult-method
#' @importFrom stats coef
setMethod(
"coefNames", "dreamletResult",
function(obj) {
unique(c(unlist(lapply(obj, function(x) colnames(coef(x))))))
}
)
#' Get error text
#'
#' Get error text
#'
#' @param obj A \code{dreamletResult} object
#'
#' @return \code{tibble} storing error text
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show errors
#' # but none are reported
#' res.err = seeErrors(res.dl)
#'
#' @rdname seeErrors-methods
#' @export
setGeneric("seeErrors", function(obj) {
standardGeneric("seeErrors")
})
#' @export
#' @rdname seeErrors-methods
#' @aliases seeErrors,dreamletResult-method
#' @importFrom dplyr as_tibble
setMethod(
"seeErrors", "dreamletResult",
function(obj) {
# Initial fit
df <- lapply(names(obj@error.initial), function(id) {
if (length(obj@error.initial[[id]]) == 0) {
return(NULL)
}
tibble(
assay = id,
errorTextInitial = obj@error.initial[[id]]
)
})
df <- bind_rows(df)
txt = paste(" Assay-level errors:", nrow(df))
message(txt)
# Gene-level
df2 <- lapply(names(obj@errors), function(id) {
if (length(obj@errors[[id]]) == 0) {
return(NULL)
}
tibble(
assay = id,
feature = names(obj@errors[[id]]),
errorText = obj@errors[[id]]
)
})
df2 <- bind_rows(df2)
txt = paste(" Gene-level errors:", nrow(df2))
message(txt)
list(assayLevel = df, geneLevel = df2)
}
)
setGeneric("assayNames", SummarizedExperiment::assayNames)
setGeneric("assay", SummarizedExperiment::assay)
# setGeneric('colData', SummarizedExperiment::colData)
# setGeneric('metadata', S4Vectors::metadata)
#' Get assayNames
#'
#' Get assayNames
#'
#' @param x dreamletResult object
#' @param ... other arguments
#'
#' @rdname assayNames-methods
#' @aliases assayNames,dreamletResult,dreamletResult-method
#' @export
setMethod(
"assayNames", signature(x = "dreamletResult"),
function(x, ...) {
names(x)
}
)
#' Get assay
#'
#' Get assay
#'
#' @param x dreamletResult object
#' @param i number indicating index, or string indicating assay
#' @param withDimnames not used
#' @param ... other arguments
#'
#' @rdname assay-methods
#' @aliases assay,dreamletResult,dreamletResult-method
#' @export
setMethod(
"assay", signature(x = "dreamletResult"),
function(x, i, withDimnames = TRUE, ...) {
x[[i]]
}
)
#' Subset with brackets
#'
#' Subset with brackets
#'
#' @param x dreamletResult object
#' @param i indeces to extract
#'
#' @rdname extract-methods
#' @aliases [,dreamletResult,dreamletResult-method
#' @export
setMethod(
"[", signature(x = "dreamletResult"),
function(x, i) {
res <- new("dreamletResult", x@.Data[i],
df_details = details(x)[i, , drop = FALSE]
)
names(res) <- names(x)[i]
res
}
)
#' Convert list of regression fits to \code{dreamletResult}
#'
#' Convert list of regression fits to \code{dreamletResult} for downstream analysis
#'
#' @param fitList list of regression fit with \code{dream()}
#' @param df_details \code{data.frame} storing assay details
#'
#' @details Useful for combining multiple runs of \code{dreamletCompareClusters()} into a single \code{dreamletResult} for downstream analysis
#' @return object of class \code{dreamletResult}
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # first comparison
#' ct.pairs <- c("B cells", "CD14+ Monocytes")
#' fit <- dreamletCompareClusters(pb, ct.pairs, method = "fixed")
#'
#' # second comparison
#' ct.pairs2 <- c("B cells", "CD8 T cells")
#' fit2 <- dreamletCompareClusters(pb, ct.pairs2, method = "fixed")
#'
#' # Make a list storing each result with a meaningful name
#' fitList <- list()
#'
#' id <- paste0("[", ct.pairs[1], "]_vs_[", ct.pairs[2], "]")
#' fitList[[id]] <- fit
#'
#' id <- paste0("[", ct.pairs2[1], "]_vs_[", ct.pairs2[2], "]")
#' fitList[[id]] <- fit2
#'
#' # create a dreamletResult form this list
#' res.compare <- as.dreamletResult(fitList)
#' res.compare
#'
#' @importFrom methods new is
#' @export
as.dreamletResult <- function(fitList, df_details = NULL) {
# check that names are not empty
if (any(names(fitList) == "")) {
stop("names(fitList) must not contain empty names")
}
# check that names are unique and not empty
if (length(unique(names(fitList))) != length(fitList)) {
stop("names(fitList) must be unique")
}
# check that entries are the result of a dream() fit
if (any(!vapply(fitList, is, class2 = "MArrayLM", FUN.VALUE = logical(1)))) {
stop("Each element of fitList must be a fit from dream()")
}
if (is.null(df_details)) {
res <- new("dreamletResult", fitList)
} else {
res <- new("dreamletResult", fitList, df_details = df_details)
}
res
}
#' Table of Top Genes from dreamlet fit
#'
#' Extract a table of the top-ranked genes from a dreamlet fit.
#'
#' @param fit dreamletResult object
#' @param coef coef
#' @param number number
#' @param genelist genelist
#' @param adjust.method adjust.method
#' @param sort.by sort.by
#' @param resort.by resort.by
#' @param p.value p.value
#' @param lfc lfc
#' @param confint confint
#'
#' @return \code{data.frame} storing hypothesis test for each gene and cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' topTable(res.dl, coef = "group_idstim", number = 3)
#'
#' @seealso \code{limma::topTable()}, \code{variancePartition::topTable()}
#' @rdname topTable-methods
#' @aliases topTable,dreamletResult,dreamletResult-method
#' @importFrom gtools smartbind
#' @export
setMethod(
"topTable", signature(fit = "dreamletResult"),
function(fit,
coef = NULL,
number = 10,
genelist = NULL,
adjust.method = "BH",
sort.by = "P",
resort.by = NULL,
p.value = 1,
lfc = 0,
confint = FALSE) {
if (any(!coef %in% coefNames(fit))) {
stop("coef must be in coefNames")
}
# Run topTable on each assay
res <- lapply(assayNames(fit), function(k) {
fit1 <- assay(fit, k)
# if entries in coef are found
if (any(coef %in% colnames(coef(fit1)))) {
# only test coefs estimated in this model
coef_found <- coef[coef %in% colnames(coef(fit1))]
# specifiying genelist create column ID
tab <- topTable(fit1, coef = coef_found, number = Inf, sort.by = "none", genelist = rownames(fit1), confint = confint)
# with multiple coefs, first column is 'ProbeID'
if (colnames(tab)[1] == "ProbeID") {
colnames(tab)[1] <- "ID"
}
if (nrow(tab) > 0) {
tab <- tab[!is.na(tab$ID), , drop = FALSE]
# if doesn't have z.std, add it
if (!"z.std" %in% colnames(tab)) {
tab$z.std <- tab$t
}
res <- data.frame(assay = k, tab)
} else {
res <- NULL
}
} else {
res <- NULL
}
res
})
# combine across assays
# allow columns to be missing when coef is array
# and some cell types dont have all of them
# Update: handle case when coef is not found
i <- which(!vapply(res, is.null, logical(1)))
if (length(i) > 1) {
res <- smartbind(list = res[i])
} else {
res <- res[i]
}
res <- DataFrame(res)
# remove rownames
rownames(res) <- c()
if (nrow(res) == 0) {
stop("No results were found matching given criteria")
}
# apply multiple testing across *all* tests
# subset based on number afterwards
res$adj.P.Val <- p.adjust(res$P.Value, adjust.method)
# apply filtering afterwards
res <- res[res$P.Value <= p.value, , drop = FALSE]
# check that logFC exists,
# since with multiple coefs it does not
if ("logFC" %in% colnames(res) & lfc != 0 & length(coef) == 1) {
res <- res[abs(res$logFC) >= lfc, , drop = FALSE]
}
if (!is.null(genelist)) {
res <- res[res$ID %in% genelist, , drop = FALSE]
}
opt <- c("logFC", "AveExpr", "P", "t", "B", "none")
if (!sort.by %in% opt) {
cmd <- paste("sort.by must be in: ", paste0(opt, collapse = ", "))
stop(cmd)
}
if( all(c("t", "F.std") %in% colnames(res)) ){
warning("Mixture of univariate and multivariate results is returned")
}
# sorting
ord <- switch(sort.by,
logFC = order(abs(res$logFC), decreasing = TRUE),
AveExpr = order(res$AveExpr, decreasing = TRUE),
P = order(res$P.Value, decreasing = FALSE),
t = order(abs(t), decreasing = TRUE),
B = order(res$B, decreasing = TRUE),
none = seq_len(nrow(res))
)
head(res[ord, ], number)
}
)
#' Test if coefficient is different from a specified value
#'
#' Test if coefficient is different from a specified value
#'
#' @param fit dreamletResult object
#' @param lfc a minimum log2-fold-change below which changes not considered scientifically meaningful
#' @param coef which coefficient to test
#' @param number number of genes to return
#' @param sort.by column to sort by
#'
#' @return \code{DataFrame} storing hypothesis test for each gene and cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' getTreat(res.dl, coef = "group_idstim", number = 3)
#'
#' @seealso \code{limma::topTreat()}, \code{variancePartition::getTreat()}
#' @importFrom variancePartition getTreat
#' @rdname getTreat-methods
#' @aliases getTreat,dreamletResult,dreamletResult-method
#' @export
setMethod(
"getTreat", signature(fit = "dreamletResult"),
function(fit, lfc = log2(1.2), coef = NULL, number = 10, sort.by = "p") {
if (any(!coef %in% coefNames(fit))) {
stop("coef must be in coefNames")
}
adjust.method <- "BH"
# Run topTable on each assay
res <- lapply(assayNames(fit), function(k) {
fit1 <- assay(fit, k)
# if coef is not found
if (all(coef %in% colnames(coef(fit1)))) {
tab <- getTreat(fit1,
lfc = lfc,
coef = coef,
number = Inf
)
res <- data.frame(assay = k, tab)
} else {
res <- NULL
}
res
})
# combine across assays
res <- DataFrame(do.call(rbind, res))
# remove rownames
rownames(res) <- c()
# apply multiple testing across *all* tests
# subset based on number afterwards
res$adj.P.Val <- p.adjust(res$P.Value, adjust.method)
opt <- c("logFC", "AveExpr", "P", "t", "B", "none")
if (!tolower(sort.by) %in% tolower(opt)) {
cmd <- paste("sort.by must be in: ", paste0(opt, collapse = ", "))
stop(cmd)
}
# sorting
ord <- switch(tolower(sort.by),
logfc = order(abs(res$logFC), decreasing = TRUE),
aveexpr = order(res$AveExpr, decreasing = TRUE),
p = order(res$P.Value, decreasing = FALSE),
t = order(abs(t), decreasing = TRUE),
b = order(res$B, decreasing = TRUE),
none = seq_len(nrow(res))
)
head(res[ord, ], number)
}
)
#' Differential expression for each assay
#'
#' Perform differential expression for each assay using linear (mixed) models
#'
#' @param x SingleCellExperiment or dreamletProcessedData object
#' @param formula regression formula for differential expression analysis
#' @param data metadata used in regression formula
#' @param assays array of assay names to include in analysis. Defaults to \code{assayNames(x)}
#' @param contrasts character vector specifying contrasts specifying linear combinations of fixed effects to test. This is fed into \code{makeContrastsDream( formula, data, contrasts=contrasts)}
#' @param min.cells minimum number of observed cells for a sample to be included in the analysis
# @param isCounts logical, indicating if data is raw counts
#' @param robust logical, use eBayes method that is robust to outlier genes
# @param normalize.method normalization method to be used by \code{calcNormFactors}
#' @param quiet show messages
#' @param BPPARAM parameters for parallel evaluation
#' @param use.eBayes should \code{eBayes} be used on result? (defualt: TRUE)
#' @param ... other arguments passed to \code{dream}
#'
#' @details
#' Fit linear (mixed) model on each cell type separately. For advanced use of contrasts see \code{variancePartition::makeContrastsDream()} and vignette \url{https://gabrielhoffman.github.io/variancePartition/articles/dream.html#advanced-hypothesis-testing-1}.
#'
#' @return Object of class \code{dreamletResult} storing results for each cell type
#'
#' @examples
#' library(muscat)
#' library(SingleCellExperiment)
#'
#' data(example_sce)
#'
#' # create pseudobulk for each sample and cell cluster
#' pb <- aggregateToPseudoBulk(example_sce,
#' assay = "counts",
#' cluster_id = "cluster_id",
#' sample_id = "sample_id",
#' verbose = FALSE
#' )
#'
#' # voom-style normalization
#' res.proc <- processAssays(pb, ~group_id)
#'
#' # Differential expression analysis within each assay,
#' # evaluated on the voom normalized data
#' res.dl <- dreamlet(res.proc, ~group_id)
#'
#' # Examine results
#' res.dl
#'
#' # Examine details for each assay
#' details(res.dl)
#'
#' # show coefficients estimated for each cell type
#' coefNames(res.dl)
#'
#' # extract results using limma-style syntax
#' # combines all cell types together
#' # adj.P.Val gives study-wide FDR
#' topTable(res.dl, coef = "group_idstim", number = 3)
#'
#' @importFrom BiocParallel SerialParam
#' @importFrom SummarizedExperiment colData assays
#' @importFrom S4Vectors as.data.frame
#' @importFrom Rdpack reprompt
#' @seealso \code{variancePartition::dream()}, \code{variancePartition::makeContrastsDream()}
#' @export
setGeneric(
"dreamlet",
function(x, formula, data = colData(x), assays = assayNames(x), contrasts = NULL, min.cells = 10, robust = FALSE, quiet = FALSE, BPPARAM = SerialParam(), use.eBayes = TRUE, ...) {
standardGeneric("dreamlet")
}
)
#' @importFrom variancePartition getContrast dream makeContrastsDream
#' @importFrom SummarizedExperiment colData assays
#' @importFrom S4Vectors as.data.frame
#' @export
#' @rdname dreamlet
#' @aliases dreamlet,dreamletProcessedData-method
setMethod(
"dreamlet", "dreamletProcessedData",
function(x, formula, data = colData(x), assays = assayNames(x), contrasts = NULL, min.cells = 10, robust = FALSE, quiet = FALSE, BPPARAM = SerialParam(), use.eBayes = TRUE, ...) {
# checks
stopifnot(is(formula, "formula"))
# check if assays are valid
if (any(!assays %in% assayNames(x))) {
idx <- which(!assays %in% assayNames(x))
txt <- paste("Assays are not found in dataset:", paste(head(assays[idx]), collapse = ", "))
stop(txt)
}
# extract metadata shared across assays
data_constant <- as.data.frame(data)
# remove samples with missing covariate data
idx <- lapply(all.vars(formula), function(v) {
which(is.na(data_constant[[v]]))
})
idx <- unique(unlist(idx))
if (length(idx) > 1) {
data_constant <- droplevels(data_constant[-idx, , drop = FALSE])
}
# for each assay
resList <- lapply(assays, function(k) {
if (!quiet) message(" ", k, "...", appendLF = FALSE)
startTime <- Sys.time()
geneExpr <- assay(x, k)
# get names of samples to extract from
# intersecting between geneExpr and metadata
ids <- intersect(colnames(geneExpr), rownames(data_constant))
geneExpr <- geneExpr[, ids, drop = FALSE]
# merge data_constant (data constant for all cell types)
# with metadata(sceObj)$aggr_means (data that varies)
data2 <- merge_metadata(
data_constant[ids, , drop = FALSE],
metadata(x),
k,
x@by
)
# drop any constant terms from the formula
form_mod <- removeConstantTerms(formula, data2)
# Drop variables in a redundant pair
form_mod <- dropRedundantTerms(form_mod, data2)
# initialize with no errors
errorArray <- NULL
# drop any constant terms from the formula
# if there is at least one variable or an intercept
tst1 = (length(all.vars(form_mod)) > 0) |
isTRUE(all.equal(form_mod, ~1))
if (tst1 & isFullRank(form_mod, data2)) {
# get contrasts customized for the formula for this cell type
if (!is.null(contrasts)) {
L <- makeContrastsDream(form_mod, data2, contrasts = contrasts, nullOnError = TRUE)
} else {
L <- NULL
}
fit <- dream(geneExpr, form_mod, data2, L = L, BPPARAM = BPPARAM, ..., hideErrorsInBackend = TRUE)
errorArray <- attr(fit, "errors")
if (!is.null(attr(fit, "error.initial"))) {
if (!quiet) message(format(Sys.time() - startTime, digits = 2))
return(list(
fit = NULL,
formula = form_mod,
n_retain = ncol(geneExpr),
errors = errorArray,
error.initial = attr(fit, "error.initial")
))
}
# if model is degenerate
if (!is.null(fit) && !any(is.na(fit$sigma))) {
if (!is.null(fit$rdf)) {
# keep genes with residual degrees of freedom > 1
# this prevents failures later
keep <- (fit$rdf >= 1)
fit <- fit[keep, ]
if (nrow(fit) == 0) fit <- NULL
}
if (use.eBayes & !is.null(fit)) {
# use counts directly if is an EList
isCounts <- ifelse(is(geneExpr, "EList"), TRUE, FALSE)
# borrow information across genes with the empirical Bayes step
fit <- eBayes(fit, robust = robust, trend = !isCounts)
}
} else {
fit <- NULL
}
} else {
fit <- NA
attr(fit, "error.initial") = "Design matrix is singular, covariates are very correlated"
}
if (!quiet) message(format(Sys.time() - startTime, digits = 2))
err.init = attr(fit, "error.initial")
if( !is(fit, "MArrayLM") && is.na(fit) ) fit <- NULL
list(
fit = fit,
formula = form_mod,
n_retain = ncol(geneExpr),
errors = errorArray,
error.initial = err.init
)
})
# name each result by the assay name
names(resList) <- assays
# extract fit
fitList <- lapply(resList, function(x) x$fit)
# only keep entries that are not NULL
# NUll is returned when coef of interest is dropped
fitList <- fitList[!vapply(fitList, is.null, FUN.VALUE = logical(1))]
# Handle errors
#--------------
# get error messages
error.initial <- lapply(resList, function(x) {
x$error.initial
})
names(error.initial) <- names(resList)
errors <- lapply(resList, function(x) {
x$errors
})
names(errors) <- names(resList)
# extract details
df_details <- lapply(names(resList), function(id) {
data.frame(
assay = id,
n_retain = resList[[id]]$n_retain,
formula = Reduce(paste, deparse(resList[[id]]$formula)),
formDropsTerms = !equalFormulas(resList[[id]]$formula, formula),
n_genes = nrow(x[[id]]), # original genes
# n_genes = nrow(resList[[id]]$fit),
n_errors = length(resList[[id]]$errors),
error_initial = ifelse(is.null(resList[[id]]$error.initial), FALSE, TRUE)
)
})
df_details <- do.call(rbind, df_details)
ndrop <- sum(df_details$formDropsTerms)
if (ndrop > 0) {
warning("Terms dropped from formulas for ", ndrop, " assays.\n Run details() on result for more information")
}
failure_frac <- sum(df_details$n_errors) / sum(df_details$n_genes)
if (failure_frac > 0) {
txt <- paste0("\nOf ", format(sum(df_details$n_genes), big.mark = ","), " models fit across all assays, ", format(failure_frac * 100, digits = 3), "% failed\n")
message(txt)
}
new("dreamletResult", fitList,
df_details = df_details,
errors = errors,
error.initial = error.initial
)
}
)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.