| 1 | 1 |
new file mode 100644 |
| ... | ... |
@@ -0,0 +1,489 @@ |
| 1 |
+#' @title Add well row and column info to rowData |
|
| 2 |
+#' |
|
| 3 |
+#' @description Add well row and column info to rowData |
|
| 4 |
+#' |
|
| 5 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 6 |
+#' |
|
| 7 |
+#' @return A \linkS4class{SummarizedExperiment} object with updated
|
|
| 8 |
+#' annotation. |
|
| 9 |
+#' |
|
| 10 |
+#' @author Jean-Philippe Fortin |
|
| 11 |
+#' |
|
| 12 |
+#' @examples |
|
| 13 |
+#' data(seExample) |
|
| 14 |
+#' seExample <- addPlateRowAndColumn(seExample) |
|
| 15 |
+#' |
|
| 16 |
+#' @importFrom stringr str_extract |
|
| 17 |
+#' @importFrom SummarizedExperiment rowData rowData<- |
|
| 18 |
+#' @export |
|
| 19 |
+addPlateRowAndColumn <- function(se){
|
|
| 20 |
+ ann <- rowData(se) |
|
| 21 |
+ if (!"Well" %in% colnames(ann)){
|
|
| 22 |
+ stop("Well must be in colData(se)")
|
|
| 23 |
+ } |
|
| 24 |
+ ann$row <- str_extract(rowData(se)[["Well"]], "[A-Z]+") |
|
| 25 |
+ ann$column <- str_extract(rowData(se)[["Well"]], "[0-9]+") |
|
| 26 |
+ ann$column <- as.integer(ann$column) |
|
| 27 |
+ rowData(se) <- ann |
|
| 28 |
+ return(se) |
|
| 29 |
+} |
|
| 30 |
+ |
|
| 31 |
+ |
|
| 32 |
+ |
|
| 33 |
+ |
|
| 34 |
+ |
|
| 35 |
+ |
|
| 36 |
+#' @title Add a column in the feature annotation for gene class |
|
| 37 |
+#' |
|
| 38 |
+#' @description Add a column in the feature annotation for gene class. |
|
| 39 |
+#' |
|
| 40 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 41 |
+#' @param gene.field String specifying the name of the column |
|
| 42 |
+#' that contains the gene names. |
|
| 43 |
+#' @param class.field String specifying the name of the column |
|
| 44 |
+#' that will store gene class. Default is "class". |
|
| 45 |
+#' @param default.class What should be the default class for |
|
| 46 |
+#' genes not specified in \code{classes}?
|
|
| 47 |
+#' Default is "target". |
|
| 48 |
+#' @param classes Named list specifying the gene classes. |
|
| 49 |
+#' Names of the list correspond to the names of the classes, |
|
| 50 |
+#' each each list element contains the gene names corresponding |
|
| 51 |
+#' the the given class. |
|
| 52 |
+#' |
|
| 53 |
+#' @return A \linkS4class{SummarizedExperiment} object with updated
|
|
| 54 |
+#' annotation. |
|
| 55 |
+#' |
|
| 56 |
+#' @author Jean-Philippe Fortin |
|
| 57 |
+#' |
|
| 58 |
+#' @importFrom stringr str_extract |
|
| 59 |
+#' @importFrom SummarizedExperiment rowData rowData<- |
|
| 60 |
+#' @export |
|
| 61 |
+addGeneClassColumn <- function(se, |
|
| 62 |
+ gene.field="Gene", |
|
| 63 |
+ class.field="class", |
|
| 64 |
+ default.class="target", |
|
| 65 |
+ classes=NULL |
|
| 66 |
+){
|
|
| 67 |
+ ann <- rowData(se) |
|
| 68 |
+ ann[[class.field]] <- default.class |
|
| 69 |
+ if (!gene.field %in% colnames(ann)){
|
|
| 70 |
+ stop("gene.field must be in colnames(rowData(se))")
|
|
| 71 |
+ } |
|
| 72 |
+ if (!is.null(classes)){
|
|
| 73 |
+ n_classes <- length(classes) |
|
| 74 |
+ for (class in names(classes)){
|
|
| 75 |
+ wh <- which(ann[[gene.field]] %in% classes[[class]]) |
|
| 76 |
+ ann[[class.field]][wh] <- class |
|
| 77 |
+ } |
|
| 78 |
+ } |
|
| 79 |
+ rowData(se) <- ann |
|
| 80 |
+ return(se) |
|
| 81 |
+} |
|
| 82 |
+ |
|
| 83 |
+ |
|
| 84 |
+ |
|
| 85 |
+ |
|
| 86 |
+#' @title log2 transformation of assay data |
|
| 87 |
+#' |
|
| 88 |
+#' @description log2 transformation of assay data. |
|
| 89 |
+#' |
|
| 90 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 91 |
+#' @param assay Integer vector specifying the indices of the assay |
|
| 92 |
+#' to log2 transform. If NULL (default), all assays |
|
| 93 |
+#' are log2 trasnformed. |
|
| 94 |
+#' @param offset Integer specifying offset to use |
|
| 95 |
+#' in the log2 transformation. |
|
| 96 |
+#' |
|
| 97 |
+#' @return A \linkS4class{SummarizedExperiment} object with
|
|
| 98 |
+#' one or many assays log2 transformed. |
|
| 99 |
+#' |
|
| 100 |
+#' @examples |
|
| 101 |
+#' data(seExample) |
|
| 102 |
+#' seExample <- logTransform(seExample, offset=1) |
|
| 103 |
+#' |
|
| 104 |
+#' @author Jean-Philippe Fortin |
|
| 105 |
+#' |
|
| 106 |
+#' @importFrom stringr str_extract |
|
| 107 |
+#' @importFrom SummarizedExperiment rowData rowData<- |
|
| 108 |
+#' @importFrom SummarizedExperiment assays assays<- |
|
| 109 |
+#' @export |
|
| 110 |
+logTransform <- function(se, |
|
| 111 |
+ assay=NULL, |
|
| 112 |
+ offset=1 |
|
| 113 |
+){
|
|
| 114 |
+ n_assays <- length(assays(se)) |
|
| 115 |
+ if (is.null(assay)){
|
|
| 116 |
+ assaysToTransform <- seq_len(n_assays) |
|
| 117 |
+ } else {
|
|
| 118 |
+ assaysToTransform <- assay |
|
| 119 |
+ } |
|
| 120 |
+ |
|
| 121 |
+ for (assay in assaysToTransform){
|
|
| 122 |
+ assays(se)[[assay]] <- log2(assays(se)[[assay]]+offset) |
|
| 123 |
+ } |
|
| 124 |
+ return(se) |
|
| 125 |
+} |
|
| 126 |
+ |
|
| 127 |
+ |
|
| 128 |
+ |
|
| 129 |
+ |
|
| 130 |
+ |
|
| 131 |
+ |
|
| 132 |
+ |
|
| 133 |
+ |
|
| 134 |
+ |
|
| 135 |
+#' @title Z score transformation using negative controls. |
|
| 136 |
+#' |
|
| 137 |
+#' @description Z score transformation using negative controls. |
|
| 138 |
+#' |
|
| 139 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 140 |
+#' @param neg.controls Character vector specifying genes that should |
|
| 141 |
+#' be used for z-score transformation. The genes must be present |
|
| 142 |
+#' in the \code{gene.field} column of \code{rowData(se)}.
|
|
| 143 |
+#' If NULL (default), all genes will be considered. |
|
| 144 |
+#' @param gene.field String specifying the name of the column |
|
| 145 |
+#' that contains the gene names specified by \code{neg.controls}.
|
|
| 146 |
+#' @param fun.centering String specifying the method |
|
| 147 |
+#' to use for centering. Either "median" (default) or "mean". |
|
| 148 |
+#' @param fun.scaling String specifying the method |
|
| 149 |
+#' to use for scaling. Either "mad" (default) or "sd". |
|
| 150 |
+#' @param scaling.ignore.controls Should the scaling factors be |
|
| 151 |
+#' calculated using all genes? TRUE by default, which |
|
| 152 |
+#' will ignore negative controls. |
|
| 153 |
+#' @param class.field String specifying the name of the column |
|
| 154 |
+#' that stored the gene class in \code{rowData(se)}.
|
|
| 155 |
+#' @param class.levels Gene classes that should be considered |
|
| 156 |
+#' for z-score transformation. If NULL (default), all classes |
|
| 157 |
+#' are considered. If specified, intersection of the specified |
|
| 158 |
+#' gene classes and the genes specified in \code{neg.controls}
|
|
| 159 |
+#' will be considered for z scoring. |
|
| 160 |
+#' |
|
| 161 |
+#' @return A \linkS4class{SummarizedExperiment} object with
|
|
| 162 |
+#' assays z-score transformed. |
|
| 163 |
+#' |
|
| 164 |
+#' @author Jean-Philippe Fortin |
|
| 165 |
+#' |
|
| 166 |
+#' @importFrom matrixStats colMedians colMads colSds |
|
| 167 |
+#' @export |
|
| 168 |
+transformToZScores <- function(se, |
|
| 169 |
+ neg.controls=NULL, |
|
| 170 |
+ gene.field="Gene", |
|
| 171 |
+ fun.centering=c("median", "mean"),
|
|
| 172 |
+ fun.scaling=c("mad", "sd"),
|
|
| 173 |
+ scaling.ignore.controls=TRUE, |
|
| 174 |
+ class.field="class", |
|
| 175 |
+ class.levels="target" |
|
| 176 |
+){
|
|
| 177 |
+ fun.centering <- match.arg(fun.centering) |
|
| 178 |
+ fun.scaling <- match.arg(fun.scaling) |
|
| 179 |
+ n_assays <- length(assays(se)) |
|
| 180 |
+ ann <- rowData(se) |
|
| 181 |
+ n_genes <- nrow(se) |
|
| 182 |
+ for (assay in seq_len(n_assays)){
|
|
| 183 |
+ Y <- assays(se)[[assay]] |
|
| 184 |
+ wh_centering <- seq_len(n_genes) |
|
| 185 |
+ wh_scaling <- seq_len(n_genes) |
|
| 186 |
+ wh_class <- seq_len(n_genes) |
|
| 187 |
+ if (!is.null(neg.controls)){
|
|
| 188 |
+ wh_centering <- which(ann[[gene.field]] %in% neg.controls) |
|
| 189 |
+ if (!scaling.ignore.controls){
|
|
| 190 |
+ wh_scaling <- which(ann[[gene.field]] %in% neg.controls) |
|
| 191 |
+ } |
|
| 192 |
+ } |
|
| 193 |
+ if (!is.null(class.levels)){
|
|
| 194 |
+ wh_class <- which(ann[[class.field]] %in% class.levels) |
|
| 195 |
+ } |
|
| 196 |
+ wh_centering <- intersect(wh_centering, wh_class) |
|
| 197 |
+ wh_scaling <- intersect(wh_scaling, wh_class) |
|
| 198 |
+ |
|
| 199 |
+ |
|
| 200 |
+ if (length(wh_centering)==0 | length(wh_scaling)==0 ){
|
|
| 201 |
+ stop("Combination of neg.controls and class.levels returns
|
|
| 202 |
+ 0 gene for transformation.") |
|
| 203 |
+ } |
|
| 204 |
+ |
|
| 205 |
+ #Centering: |
|
| 206 |
+ if (fun.centering=="median"){
|
|
| 207 |
+ factors <- colMedians(Y[wh_centering,,drop=FALSE], na.rm=TRUE) |
|
| 208 |
+ } else {
|
|
| 209 |
+ factors <- colMeans(Y[wh_centering,,drop=FALSE], na.rm=TRUE) |
|
| 210 |
+ } |
|
| 211 |
+ Y <- sweep(Y, 2 ,factors, "-") |
|
| 212 |
+ |
|
| 213 |
+ #Scaling: |
|
| 214 |
+ if (fun.scaling=="mad"){
|
|
| 215 |
+ factors <- colMads(Y[wh_scaling,,drop=FALSE], na.rm=TRUE) |
|
| 216 |
+ } else {
|
|
| 217 |
+ factors <- colSds(Y[wh_scaling,,drop=FALSE], na.rm=TRUE) |
|
| 218 |
+ } |
|
| 219 |
+ Y <- sweep(Y, 2 ,factors, "/") |
|
| 220 |
+ assays(se)[[assay]] <- Y |
|
| 221 |
+ } |
|
| 222 |
+ return(se) |
|
| 223 |
+} |
|
| 224 |
+ |
|
| 225 |
+ |
|
| 226 |
+ |
|
| 227 |
+#' @title Calculate limma statistics |
|
| 228 |
+#' |
|
| 229 |
+#' @description Calculate limma statistics. |
|
| 230 |
+#' |
|
| 231 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 232 |
+#' @param design Design matrix to pass to limma. |
|
| 233 |
+#' @param contrast.matrix Contrast matrix to pass to limma. |
|
| 234 |
+#' |
|
| 235 |
+#' @return For each assay in \code{se}, it returns a named
|
|
| 236 |
+#' list with the following elements: |
|
| 237 |
+#' - lfc: log-fold changes |
|
| 238 |
+#' - pval: p-lvaues |
|
| 239 |
+#' - fdr: FDR-adjusted p-values. |
|
| 240 |
+#' |
|
| 241 |
+#' |
|
| 242 |
+#' @examples |
|
| 243 |
+#' data(seExample) |
|
| 244 |
+#' library(SummarizedExperiment) |
|
| 245 |
+#' |
|
| 246 |
+#' classes <- list(neg=c("OR5M9","OR6N1"),
|
|
| 247 |
+#' pos=c("KRAS","NRAS"),
|
|
| 248 |
+#' rnaimax=c("RNAiMAX"),
|
|
| 249 |
+#' empty="empty") |
|
| 250 |
+#' seExample <- addGeneClassColumn(seExample, classes=classes) |
|
| 251 |
+#' |
|
| 252 |
+#' # Transforming the data: |
|
| 253 |
+#' seExample <- logTransform(seExample, offset=1) |
|
| 254 |
+#' seExample <- normalizeBetweenPlates(seExample) |
|
| 255 |
+#' seExample <- transformToZScores(seExample) |
|
| 256 |
+#' |
|
| 257 |
+#' # Creating design matrix: |
|
| 258 |
+#' pheno <- as.factor(colData(seExample)$Condition) |
|
| 259 |
+#' pheno <- relevel(pheno, ref="WT") |
|
| 260 |
+#' design <- model.matrix(~pheno-1) |
|
| 261 |
+#' colnames(design) <- gsub("pheno", "", colnames(design))
|
|
| 262 |
+#' contrasts <- c(Mutant = "Mutant", |
|
| 263 |
+#' Wildtype="WT", |
|
| 264 |
+#' Mutant_vs_Wildtype="Mutant - WT") |
|
| 265 |
+#' contrast.matrix <- cbind("Wildtype"=c(1,0),
|
|
| 266 |
+#' "Mutant"=c(0,1), |
|
| 267 |
+#' "Mutant_vs_Wildtype"=c(-1,1)) |
|
| 268 |
+#' |
|
| 269 |
+#' # Calculate statistics: |
|
| 270 |
+#' stats <- calculateStatistics(seExample, |
|
| 271 |
+#' design=design, |
|
| 272 |
+#' contrast.matrix=contrast.matrix) |
|
| 273 |
+#' |
|
| 274 |
+#' @author Jean-Philippe Fortin |
|
| 275 |
+#' |
|
| 276 |
+#' @importFrom limma lmFit contrasts.fit eBayes |
|
| 277 |
+#' @importFrom stats p.adjust |
|
| 278 |
+#' @export |
|
| 279 |
+calculateStatistics <- function(se, |
|
| 280 |
+ design, |
|
| 281 |
+ contrast.matrix=NULL |
|
| 282 |
+){
|
|
| 283 |
+ n_assays <- length(assays(se)) |
|
| 284 |
+ results <- list() |
|
| 285 |
+ for (assay in seq_len(n_assays)){
|
|
| 286 |
+ Y <- assays(se)[[assay]] |
|
| 287 |
+ fit <- lmFit(Y, design) |
|
| 288 |
+ if (!is.null(contrast.matrix)){
|
|
| 289 |
+ fit <- contrasts.fit(fit, contrast.matrix) |
|
| 290 |
+ } |
|
| 291 |
+ fit <- eBayes(fit) |
|
| 292 |
+ stats <- list() |
|
| 293 |
+ stats[["lfc"]] <- fit$coefficients |
|
| 294 |
+ stats[["pval"]] <- fit$p.value |
|
| 295 |
+ stats[["fdr"]] <- apply(stats[["pval"]], 2, p.adjust, "fdr") |
|
| 296 |
+ results[[assay]] <- stats |
|
| 297 |
+ } |
|
| 298 |
+ names(results) <- names(assays(se)) |
|
| 299 |
+ return(results) |
|
| 300 |
+} |
|
| 301 |
+ |
|
| 302 |
+ |
|
| 303 |
+ |
|
| 304 |
+#' @title Between-plate normalization |
|
| 305 |
+#' |
|
| 306 |
+#' @description Between-plate normalization. |
|
| 307 |
+#' |
|
| 308 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 309 |
+#' @param neg.controls Character vector specifying genes that should |
|
| 310 |
+#' be used for estimating normalizatin factors. |
|
| 311 |
+#' The genes must be present in the \code{gene.field}
|
|
| 312 |
+#' column of \code{rowData(se)}.
|
|
| 313 |
+#' If NULL (default), all genes will be considered. |
|
| 314 |
+#' @param fun String specifying the method |
|
| 315 |
+#' to use for normalization. Either "median" (default) or "mean". |
|
| 316 |
+#' @param plate.field String specifying the name of the column |
|
| 317 |
+#' that contains the plate ID. |
|
| 318 |
+#' @param gene.field String specifying the name of the column |
|
| 319 |
+#' that contains the gene names specified by \code{neg.controls}.
|
|
| 320 |
+#' @param class.field String specifying the name of the column |
|
| 321 |
+#' that stored the gene class in \code{rowData(se)}.
|
|
| 322 |
+#' @param class.levels Gene classes that should be considered |
|
| 323 |
+#' for estimating normalization factors. If NULL (default), all classes |
|
| 324 |
+#' are considered. If specified, intersection of the specified |
|
| 325 |
+#' gene classes and the genes specified in \code{neg.controls}
|
|
| 326 |
+#' will be considered for z scoring. |
|
| 327 |
+#' |
|
| 328 |
+#' @return A \linkS4class{SummarizedExperiment} object with
|
|
| 329 |
+#' value normalized between plates. |
|
| 330 |
+#' |
|
| 331 |
+#' @examples |
|
| 332 |
+#' data(seExample) |
|
| 333 |
+#' seExample <- logTransform(seExample, offset=1) |
|
| 334 |
+#' |
|
| 335 |
+#' classes <- list(neg=c("OR5M9","OR6N1"),
|
|
| 336 |
+#' pos=c("KRAS","NRAS"),
|
|
| 337 |
+#' rnaimax=c("RNAiMAX"),
|
|
| 338 |
+#' empty="empty") |
|
| 339 |
+#' seExample <- addGeneClassColumn(seExample, classes=classes) |
|
| 340 |
+#' |
|
| 341 |
+#' seExample <- normalizeBetweenPlates(seExample) |
|
| 342 |
+#' |
|
| 343 |
+#' @author Jean-Philippe Fortin |
|
| 344 |
+#' |
|
| 345 |
+#' @importFrom matrixStats colMedians |
|
| 346 |
+#' @importFrom SummarizedExperiment rowData assays assays<- |
|
| 347 |
+#' @export |
|
| 348 |
+normalizeBetweenPlates <- function(se, |
|
| 349 |
+ neg.controls=NULL, |
|
| 350 |
+ fun=c("median", "mean"),
|
|
| 351 |
+ plate.field="Plate", |
|
| 352 |
+ gene.field="Gene", |
|
| 353 |
+ class.field="class", |
|
| 354 |
+ class.levels=c("target")
|
|
| 355 |
+){
|
|
| 356 |
+ fun <- match.arg(fun) |
|
| 357 |
+ n_assays <- length(assays(se)) |
|
| 358 |
+ assaysToNormalize <- seq_len(n_assays) |
|
| 359 |
+ ann <- rowData(se) |
|
| 360 |
+ plate <- ann[[plate.field]] |
|
| 361 |
+ plates <- unique(plate) |
|
| 362 |
+ n_plates <- length(plates) |
|
| 363 |
+ n_genes <- nrow(se) |
|
| 364 |
+ if (!gene.field %in% colnames(ann)){
|
|
| 365 |
+ stop("gene.field not found in colnames(rowData(se))")
|
|
| 366 |
+ } |
|
| 367 |
+ if (!class.field %in% colnames(ann)){
|
|
| 368 |
+ stop("class.field not found in colnames(rowData(se))")
|
|
| 369 |
+ } |
|
| 370 |
+ for (assay in assaysToNormalize){
|
|
| 371 |
+ Y <- assays(se)[[assay]] |
|
| 372 |
+ for (k in seq_len(n_plates)){
|
|
| 373 |
+ if (is.null(neg.controls)){
|
|
| 374 |
+ wh_control <- seq_len(n_genes) |
|
| 375 |
+ } else {
|
|
| 376 |
+ wh_control <- which(ann[[gene.field]] %in% neg.controls) |
|
| 377 |
+ } |
|
| 378 |
+ wh_plate <- which(plate==plates[[k]]) |
|
| 379 |
+ wh_norm <- intersect(wh_control, wh_plate) |
|
| 380 |
+ |
|
| 381 |
+ # Only normalizing using target genes: |
|
| 382 |
+ if (!is.null(class.levels)){
|
|
| 383 |
+ wh_class <- which(ann[[class.field]] %in% class.levels) |
|
| 384 |
+ } else {
|
|
| 385 |
+ wh_class <- seq_len(n_genes) |
|
| 386 |
+ } |
|
| 387 |
+ wh_norm <- intersect(wh_norm, wh_class) |
|
| 388 |
+ if (fun=="median"){
|
|
| 389 |
+ factors <- colMedians(Y[wh_norm,,drop=FALSE], na.rm=TRUE) |
|
| 390 |
+ } else {
|
|
| 391 |
+ factors <- colMeans(Y[wh_norm,,drop=FALSE], na.rm=TRUE) |
|
| 392 |
+ } |
|
| 393 |
+ Y[wh_plate,] <- sweep(Y[wh_plate,,drop=FALSE], 2 ,factors, "-") |
|
| 394 |
+ } |
|
| 395 |
+ assays(se)[[assay]] <- Y |
|
| 396 |
+ } |
|
| 397 |
+ return(se) |
|
| 398 |
+} |
|
| 399 |
+ |
|
| 400 |
+ |
|
| 401 |
+ |
|
| 402 |
+#' @title Within-plate normalization |
|
| 403 |
+#' |
|
| 404 |
+#' @description Within-plate normalization |
|
| 405 |
+#' |
|
| 406 |
+#' @param se A \linkS4class{SummarizedExperiment} object.
|
|
| 407 |
+#' @param fun String specifying the method |
|
| 408 |
+#' to use for normalization. Either "median" (default) or "mean". |
|
| 409 |
+#' @param plate.field String specifying the name of the column |
|
| 410 |
+#' that contains the plate ID. |
|
| 411 |
+#' @param what String specifying if plate rows or columns |
|
| 412 |
+#' should be normalized. "row" by default. |
|
| 413 |
+#' @param class.field String specifying the name of the column |
|
| 414 |
+#' that stored the gene class in \code{rowData(se)}.
|
|
| 415 |
+#' @param class.levels Gene classes that should be considered |
|
| 416 |
+#' for estimating normalization factors. If NULL (default), all classes |
|
| 417 |
+#' are considered. If specified, intersection of the specified |
|
| 418 |
+#' gene classes and the genes specified in \code{neg.controls}
|
|
| 419 |
+#' will be considered for z scoring. |
|
| 420 |
+#' |
|
| 421 |
+#' @return A \linkS4class{SummarizedExperiment} object with
|
|
| 422 |
+#' value normalized within plates. |
|
| 423 |
+#' |
|
| 424 |
+#' #' @examples |
|
| 425 |
+#' data(seExample) |
|
| 426 |
+#' seExample <- logTransform(seExample, offset=1) |
|
| 427 |
+#' seExample <- normalizeWithinPlate(seExample) |
|
| 428 |
+#' |
|
| 429 |
+#' @author Jean-Philippe Fortin |
|
| 430 |
+#' @importFrom SummarizedExperiment rowData assays assays<- |
|
| 431 |
+#' @importFrom matrixStats colMedians |
|
| 432 |
+#' @export |
|
| 433 |
+normalizeWithinPlate <- function(se, |
|
| 434 |
+ fun=c("median", "mean"),
|
|
| 435 |
+ what=c("row", "column"),
|
|
| 436 |
+ plate.field="Plate", |
|
| 437 |
+ class.field="class", |
|
| 438 |
+ class.levels=c("target")
|
|
| 439 |
+){
|
|
| 440 |
+ fun <- match.arg(fun) |
|
| 441 |
+ what <- match.arg(what) |
|
| 442 |
+ n_assays <- length(assays(se)) |
|
| 443 |
+ ann <- rowData(se) |
|
| 444 |
+ if (!what %in% colnames(ann)){
|
|
| 445 |
+ se <- addPlateRowAndColumn(se) |
|
| 446 |
+ ann <- rowData(se) |
|
| 447 |
+ } |
|
| 448 |
+ plate <- ann[[plate.field]] |
|
| 449 |
+ slice <- paste0(plate, "_", ann[[what]]) |
|
| 450 |
+ slices <- unique(slice) |
|
| 451 |
+ n_slices <- length(slices) |
|
| 452 |
+ n_genes <- nrow(se) |
|
| 453 |
+ for (assay in seq_len(n_assays)){
|
|
| 454 |
+ Y <- assays(se)[[assay]] |
|
| 455 |
+ for (k in seq_len(n_slices)){
|
|
| 456 |
+ wh_control <- seq_len(n_genes) |
|
| 457 |
+ wh_slice <- which(slice==slices[[k]]) |
|
| 458 |
+ wh_norm <- intersect(wh_control, wh_slice) |
|
| 459 |
+ # Only normalizing using target genes: |
|
| 460 |
+ if (!is.null(class.levels)){
|
|
| 461 |
+ wh_class <- which(ann[[class.field]] %in% class.levels) |
|
| 462 |
+ } else {
|
|
| 463 |
+ wh_class <- seq_len(n_genes) |
|
| 464 |
+ } |
|
| 465 |
+ wh_norm <- intersect(wh_norm, wh_class) |
|
| 466 |
+ if (fun=="median"){
|
|
| 467 |
+ factors <- colMedians(Y[wh_norm,,drop=FALSE], na.rm=TRUE) |
|
| 468 |
+ } else {
|
|
| 469 |
+ factors <- colMeans(Y[wh_norm,,drop=FALSE], na.rm=TRUE) |
|
| 470 |
+ } |
|
| 471 |
+ Y[wh_slice,] <- sweep(Y[wh_slice,,drop=FALSE], 2 ,factors, "-") |
|
| 472 |
+ } |
|
| 473 |
+ assays(se)[[assay]] <- Y |
|
| 474 |
+ } |
|
| 475 |
+ return(se) |
|
| 476 |
+} |
|
| 477 |
+ |
|
| 478 |
+ |
|
| 479 |
+ |
|
| 480 |
+ |
|
| 481 |
+ |
|
| 482 |
+ |
|
| 483 |
+ |
|
| 484 |
+ |
|
| 485 |
+ |
|
| 486 |
+ |
|
| 487 |
+ |
|
| 488 |
+ |
|
| 489 |
+ |
