R - 如何根据列和列名重塑数据框？

Question

我需要根据 colnames 和每列的值重塑数据框。如何使用 reshape（或任何其他库，例如 dplyr）来实现这一点？

我相信melt功能会很有用，但我不是很熟悉。

df <- data.frame(
  Group1 = c("EMX1", "EXO_C3L4", "FAF2P1", "FAM224A","GATC", 
             "FAM43A", "FAT4", "EXO_FEZF1-AS1"),
  AVG_EXPR = rnorm(8),
  Group2 = c("EXO_BRPF3", "AFS", "IJAS", "CCDC187", 
             "CCDC200", "CCDC7", "CCL27", "CD6"),
  AVG_EXPR = rnorm(8))
> df
         Group1   AVG_EXPR    Group2 AVG_EXPR.1
1          EMX1 -0.4792520 EXO_BRPF3 -0.8802157
2      EXO_C3L4 -0.4820114       AFS -0.2135738
3        FAF2P1  1.0326328      IJAS -0.6290447
4       FAM224A  1.1318738   CCDC187  1.9813584
5          GATC -0.5019679   CCDC200  0.6749088
6        FAM43A -2.0718576     CCDC7 -0.4217126
7          FAT4 -0.1194380     CCL27 -0.3552010
8 EXO_FEZF1-AS1  1.4106888       CD6  0.4217045

所以我的新 data.frame 看起来像：

df2 <- data.frame(
  Gene = c("EMX1", "EXO_C3L4", "FAF2P1", "FAM224A","GATC", 
           "FAM43A", "FAT4", "EXO_FEZF1-AS1","EXO_BRPF3", "AFS", "IJAS", "CCDC187", 
           "CCDC200", "CCDC7", "CCL27", "CD6"),
  AVG_EXPR = c(df$AVG_EXPR,df$AVG_EXPR.1),
  Group = c(rep("Group1",8),rep("Group2",8))
)

> df2
            Gene   AVG_EXPR  Group
1           EMX1 -0.4792520 Group1
2       EXO_C3L4 -0.4820114 Group1
3         FAF2P1  1.0326328 Group1
4        FAM224A  1.1318738 Group1
5           GATC -0.5019679 Group1
6         FAM43A -2.0718576 Group1
7           FAT4 -0.1194380 Group1
8  EXO_FEZF1-AS1  1.4106888 Group1
9      EXO_BRPF3 -0.8802157 Group2
10           AFS -0.2135738 Group2
11          IJAS -0.6290447 Group2
12       CCDC187  1.9813584 Group2
13       CCDC200  0.6749088 Group2
14         CCDC7 -0.4217126 Group2
15         CCL27 -0.3552010 Group2
16           CD6  0.4217045 Group2

Answer 1

一个选项是pivot_longer 的开发版tidyr

library(tidyr)
library(dplyr)
library(tibble)
nm1 <- sub("\\.?\\d+$", "", names(df))
names(df) <- paste0(nm1, ":", ave(seq_along(nm1), nm1, FUN = seq_along))
df %>% 
    rownames_to_column('rn') %>% 
    pivot_longer(-rn, names_to= c(".value", "Group"), names_sep= ":")

Answer 2

您可以将两个dplyr 调用与bind_rows 结合起来。

df %>% select(gene=Group1,AVG_EXPR) %>% mutate(group='group 1') %>% 
  bind_rows(.,df %>% select(gene=Group2,AVG_EXPR=AVG_EXPR.1) %>% mutate(group='group 2'))

Answer 3

您可以使用基本 R 的 reshape。诀窍是嵌套varying 和cbind 一个id 列。

reshape(cbind(id=1:nrow(DF), DF), varying=list(c(2, 4), c(3, 5)), direction="long", 
        timevar="Group", v.names=c("Gene", "AVG_EXPR"), times=c("Group1", "Group2"))
#          id  Group          Gene    AVG_EXPR
# 1.Group1  1 Group1          EMX1  0.55748718
# 2.Group1  2 Group1      EXO_C3L4 -0.71399573
# 3.Group1  3 Group1        FAF2P1  0.72595733
# 4.Group1  4 Group1       FAM224A  0.78106399
# 5.Group1  5 Group1          GATC  0.33296728
# 6.Group1  6 Group1        FAM43A -1.14699031
# 7.Group1  7 Group1          FAT4  0.64475462
# 8.Group1  8 Group1 EXO_FEZF1-AS1  0.88621529
# 1.Group2  1 Group2     EXO_BRPF3 -0.62445044
# 2.Group2  2 Group2           AFS  0.14577195
# 3.Group2  3 Group2          IJAS -1.15083794
# 4.Group2  4 Group2       CCDC187  0.95483371
# 5.Group2  5 Group2       CCDC200 -1.55217630
# 6.Group2  6 Group2         CCDC7 -0.51225644
# 7.Group2  7 Group2         CCL27  0.52270643
# 8.Group2  8 Group2           CD6 -0.08751736

数据

DF <- structure(list(Group1 = structure(c(1L, 2L, 4L, 5L, 8L, 6L, 7L, 
3L), .Label = c("EMX1", "EXO_C3L4", "EXO_FEZF1-AS1", "FAF2P1", 
"FAM224A", "FAM43A", "FAT4", "GATC"), class = "factor"), AVG_EXPR = c(0.557487175664929, 
-0.713995729705457, 0.725957334982896, 0.781063988053138, 0.332967281017435, 
-1.14699031084438, 0.644754618475526, 0.88621528791546), Group2 = structure(c(7L, 
1L, 8L, 2L, 3L, 4L, 5L, 6L), .Label = c("AFS", "CCDC187", "CCDC200", 
"CCDC7", "CCL27", "CD6", "EXO_BRPF3", "IJAS"), class = "factor"), 
    AVG_EXPR.1 = c(-0.624450436709626, 0.145771950049532, -1.15083793547685, 
    0.954833708527147, -1.55217630379434, -0.512256436764118, 
    0.522706429197282, -0.0875173629601027)), class = "data.frame", row.names = c(NA, 
-8L))