同名公主號(hào)：BBio

reshape2包最常用的場(chǎng)景就是長(zhǎng)數(shù)據(jù)和寬數(shù)據(jù)的轉(zhuǎn)換，兩個(gè)函數(shù)即可搞定。作者也是tidyverse的作者，Rstudio項(xiàng)目的首席科學(xué)家Hadley Wickham。reshape2已經(jīng)很久沒(méi)有更新了，tidyverse工作流下的tidyr是實(shí)現(xiàn)數(shù)據(jù)轉(zhuǎn)換更好的工具，不過(guò)此處還是簡(jiǎn)單介紹一下。

//長(zhǎng)數(shù)據(jù)：ggplot2繪圖數(shù)據(jù)格式

data <- data.frame(celltype=rep(c("T cells", "B cells"), each=3), 
                   sample=rep(c("S1", "S2", "S3"), 2), 
                  percent=runif(6))
#celltype sample   percent
#1  T cells     S1 0.9825346
#2  T cells     S2 0.3768486
#3  T cells     S3 0.8072078
#4  B cells     S1 0.8889155
#5  B cells     S2 0.8150424
#6  B cells     S3 0.5531544

//寬數(shù)據(jù)：pheatmap繪圖數(shù)據(jù)格式

data <- data.frame(celltype=c("T cells", "B cells"), 
                  S1=runif(2), 
                  S2=runif(2), 
                  S3=runif(2))
#celltype        S1        S2         S3
#1  T cells 0.1704449 0.3704776 0.82863072
#2  B cells 0.6085021 0.5197027 0.09653481

//如何實(shí)現(xiàn)兩者的相互轉(zhuǎn)換？

• melt函數(shù)：寬數(shù)據(jù)轉(zhuǎn)為長(zhǎng)數(shù)據(jù)

library(reshape2)
?melt

data <- data.frame(celltype=c("T cells", "B cells"), 
                  S1=runif(2), 
                  S2=runif(2), 
                  S3=runif(2))

#id.vars定義錨點(diǎn)，也可以直接寫id
melt(data, id.vars=c("celltype"), variable.name="sample", value.name="percent")
#celltype sample    percent
#1  T cells     S1 0.17044494
#2  B cells     S1 0.60850210
#3  T cells     S2 0.37047760
#4  B cells     S2 0.51970266
#5  T cells     S3 0.82863072
#6  B cells     S3 0.09653481

• dcast函數(shù)：長(zhǎng)數(shù)據(jù)轉(zhuǎn)為寬數(shù)據(jù)

?dcast

data <- data.frame(celltype=rep(c("T cells", "B cells"), each=3), 
                   sample=rep(c("S1", "S2", "S3"), 2), 
                  percent=runif(6))
dcast(data, celltype ~ sample, value.var="percent")

#如果數(shù)據(jù)有多余的列怎么辦？指定value.var定義解析為值的列
#如果解析的行有多個(gè)值怎么辦？添加運(yùn)算函數(shù)
names(airquality) <- tolower(names(airquality))
aqm <- melt(airquality, id=c("month", "day"), na.rm=TRUE)
head(aqm)
#  month day variable value
#1     5   1    ozone    41
#2     5   2    ozone    36
#3     5   3    ozone    12
#4     5   4    ozone    18
#6     5   6    ozone    28
#7     5   7    ozone    23

#month和day對(duì)應(yīng)的variable只有一行，不需要指定運(yùn)算函數(shù)
dcast(aqm, month + day ~ variable)

#month和variable有多個(gè)行，指定運(yùn)算函數(shù)，不然會(huì)統(tǒng)計(jì)頻數(shù)
dcast(aqm, month ~ variable, fun.aggregate=mean)