重塑然后折叠数据框

Question

努力将我凌乱、长度不等的data.frame 从宽表转换为长表，然后折叠（汇总）新变量。目前看起来是这样的，Gene 作为一个变量，GO_terms 作为一个包含多个逗号分隔值的变量：

Gene            GO_terms                        
AA1006G00001    GO:0098655, GO:0008643, GO:0005351, GO:0005886, GO:0016021      
AA100G00001     GO:0098655, GO:0009944, GO:0009862, GO:0010075, GO:0010014, GO:0009855, GO:0010310
AA100G00002     GO:0098655, GO:0008643, GO:0005886

我要做的第一步是转换为“长”格式，所以它看起来像这样：

Gene            GO_terms 
AA1006G00001    GO:0098655
AA1006G00001    GO:0008643
AA1006G00001    GO:0005351
AA1006G00001    GO:0005886
AA1006G00001    GO:0016021
AA100G00001     GO:0001666
AA100G00001     GO:0009944
AA100G00001     GO:0009862
AA100G00001     GO:0010075
AA100G00001     GO:0010014
AA100G00001     GO:0009855
AA100G00001     GO:0010310
AA100G00002     GO:0008270
AA100G00002     GO:0005634
AA100G00002     GO:0005886
AA100G00003     GO:0005488
AA100G00003     GO:0005634

然后，我希望通过切换两个变量来重新组织这个data.table，因为它的整理如下：

GO_terms    Genes
GO:0005351  AA1006G00001        
GO:0005886  AA1006G00001,   AA100G00002 
GO:0008643  AA1006G00001,   AA100G00002 
GO:0009855  AA100G00001     
GO:0009862  AA100G00001     
GO:0009944  AA100G00001     
GO:0010014  AA100G00001     
GO:0010075  AA100G00001     
GO:0010310  AA100G00001     
GO:0016021  AA1006G00001        
GO:0098655  AA1006G00001,   AA100G00001,      AA100G00002

包含基因的变量可以在一列中（使用逗号分隔值），也可以在多列中。

谁能提供tidyr、reshape2 或dplyr 解决方案？

编辑：dput() 表是：

structure(list(`Gene    ` = c("AA1006G00001\t", "AA100G00001\t", 
"AA100G00002\t"), `GO_terms                     ` = c("GO:0098655, GO:0008643, GO:0005351, GO:0005886, GO:0016021\t\t", 
"GO:0098655, GO:0009944, GO:0009862, GO:0010075, GO:0010014, GO:0009855, GO:0010310", 
"GO:0098655, GO:0008643, GO:0005886")), row.names = c(NA, -3L
), class = c("tbl_df", "tbl", "data.frame"), spec = structure(list(
    cols = list(`Gene   ` = structure(list(), class = c("collector_character", 
    "collector")), `GO_terms                        ` = structure(list(), class = c("collector_character", 
    "collector"))), default = structure(list(), class = c("collector_guess", 
    "collector"))), class = "col_spec"))

Answer 1

看来您正在做一些 GO 分析。您可以从topGO（Bioconductor 中最流行的 GO 分析 R 包之一）尝试inverseList：

library(topGO)

gene.to.go <- strsplit(gsub('\t', '', df$GO_terms), ', ', fixed = TRUE)
names(gene.to.go) <- gsub('\t', '', df$Gene)

go.to.gene <- inverseList(gene.to.go)

data.frame(GO_term = names(go.to.gene), Genes = sapply(go.to.gene, paste0, collapse = ', '),
           stringsAsFactors = FALSE, row.names = NULL)

#       GO_term                                  Genes
# 1  GO:0005351                           AA1006G00001
# 2  GO:0005886              AA1006G00001, AA100G00002
# 3  GO:0008643              AA1006G00001, AA100G00002
# 4  GO:0009855                            AA100G00001
# 5  GO:0009862                            AA100G00001
# 6  GO:0009944                            AA100G00001
# 7  GO:0010014                            AA100G00001
# 8  GO:0010075                            AA100G00001
# 9  GO:0010310                            AA100G00001
# 10 GO:0016021                           AA1006G00001
# 11 GO:0098655 AA1006G00001, AA100G00001, AA100G00002

其实在topGO中导入带有readMappings的GO映射文件会更容易对数据进行操作。

Answer 2

这是一个 tidyr 和 dplyr 的解决方案：

library(tidyr)
library(dplyr)

#allow up to seven Genes per GO_term if there is more increase the letters expression
long<-df %>% separate(GO_terms, into=paste0("a", 1:100), sep=", ", extra="merge") %>% 
  gather( key="key", value="GO_terms", -Gene)

#filter data frame, remove the NA and keep the desired columns
long<-long[!is.na(long$GO_terms), c("Gene", "GO_terms")]

final<-long %>% group_by(GO_terms) %>% summarize( Gene=toString(Gene) )