数据帧处理

Question

我有一个由Match <- read.table("Match.txt", sep="", fill =T, stringsAsFactors = FALSE, quote = "", header = F) 读取的数据框，看起来像这样：

> ab
           V1       V2  V3                       V4 V5    V6 V7    V8 V9               V10
1  Inspecting sequence  ID chr1:173244300-173244500       NA       NA                     
2   V$ATF3_Q6        |  19                      (-)  | 0.877  | 0.622  |    aagtccCATCAggg
3   V$ATF3_Q6        |  34                      (-)  | 0.788  | 0.655  |    agggaaCGACAcag
4   V$ATF3_Q6        | 102                      (+)  | 0.738  | 0.685  |    cccTGAGCttagga
5  V$CEBPB_01        |  24                      (+)  | 0.950  | 0.882  |    ccatcagGGAAGgg
72   V$YY1_01        | 117                      (+)  | 0.996  | 0.984  | acttCCCATcttttaag
73 Inspecting sequence  ID chr1:173244350-173244550       NA       NA                     
74  V$ATF3_Q6        |  52                      (+)  | 0.738  | 0.685  |    cccTGAGCttagga
75  V$ATF3_Q6        | 160                      (+)  | 0.862  | 0.687  |    gtcTGACCtggaga
76 V$CEBPB_01        |  57                      (+)  | 0.966  | 0.958  |    agcttagGAAACtt

它包含百万个这样的重复，其中第一行是：Inspecting sequence ID chr1:173244300-173244500，然后是上面可以看到的一些值。我想处理它，记住以下几点：

1. 提取第一行，在: 和- 上断开它，这样我将得到三列，例如：chr1 173244300 173244500
2. 第 4 列应包含 V1$Row2 第一个元素，在 $ 和 _ 上拆分，然后取第二个索引，即 ATF3，像这样我有 30 个确定的（让我们称之为名称）案例，在每种情况下都会观察到一些，而另一些则不会（1 个案例是从第 1 行到第 72 行，第二个是从第 73 行开始）。
3. 如果该名称出现在 1 个案例中，则将值 B 分配给该列，否则将分配值 U

所以根据我的输入，我想得到以下输出：

chr     start       stop        ATF3  CEBPB  YY1    ..(All which appear e.g from row 1 to 72, ignoring duplicates)
chr1    173244300   173244500   B     B      B  
chr1    173244350   173244550   B     B      U

我想在标题中设置一个固定编号列（我知道它们是 32 个这样的名称），因此如果它们出现在一种情况下，将分配 B，否则将分配 U。

如果有人能帮我做这件事，那将是一个很大的帮助。

这是此示例数据帧的输出：

> ab <- dput(Match[c(1:5,72:76), ])
structure(list(V1 = c("Inspecting", "V$ATF3_Q6", "V$ATF3_Q6", 
"V$ATF3_Q6", "V$CEBPB_01", "V$YY1_01", "Inspecting", "V$ATF3_Q6", 
"V$ATF3_Q6", "V$CEBPB_01"), V2 = c("sequence", "|", "|", "|", 
"|", "|", "sequence", "|", "|", "|"), V3 = c("ID", "19", "34", 
"102", "24", "117", "ID", "52", "160", "57"), V4 = c("chr1:173244300-173244500", 
"(-)", "(-)", "(+)", "(+)", "(+)", "chr1:173244350-173244550", 
"(+)", "(+)", "(+)"), V5 = c("", "|", "|", "|", "|", "|", "", 
"|", "|", "|"), V6 = c(NA, 0.877, 0.788, 0.738, 0.95, 0.996, 
NA, 0.738, 0.862, 0.966), V7 = c("", "|", "|", "|", "|", "|", 
"", "|", "|", "|"), V8 = c(NA, 0.622, 0.655, 0.685, 0.882, 0.984, 
NA, 0.685, 0.687, 0.958), V9 = c("", "|", "|", "|", "|", "|", 
"", "|", "|", "|"), V10 = c("", "aagtccCATCAggg", "agggaaCGACAcag", 
"cccTGAGCttagga", "ccatcagGGAAGgg", "acttCCCATcttttaag", "", 
"cccTGAGCttagga", "gtcTGACCtggaga", "agcttagGAAACtt")), .Names = c("V1", 
"V2", "V3", "V4", "V5", "V6", "V7", "V8", "V9", "V10"), row.names = c(1L, 
2L, 3L, 4L, 5L, 72L, 73L, 74L, 75L, 76L), class = "data.frame")

Answer 1

将this question 中的输入文件设为/c/tmp.txt

这个 awk 脚本保存为 SO-38563400.awk:

BEGIN {
 OFS="\t" # Set the output separator
 i=0 # Just to init the counter and be sure to start at 1 later
}
 {
 #print $0
 }
/Inspecting sequence ID/ { # Changing sequence, initialize new entry with start and end
  split($4,arr,"[:-]") # split the string in fields, split on : and -
  seq[i++,"chr"]=arr[1] # Save the chr part and increase the sequence beforehand
  seq[i,"start"]=arr[2] # save the start date
  seq[i,"end"]=arr[3] # Save the end date
}

/V[$][^_]+_.*/ { # V line type,
  split($1,arr,"[$_]") # Split on $ and underscore
  seq[i,arr[2]]="B" # This has been seen, setting to B
  seq[i,"print"]=1
  names[arr[2]]++ # Save the name for output
  # (and count occurences, just for fun, well mainly because an int is cheaper to store)
  # Main reason is it allow a quicker access toa rray keys ant END block
}

END {
  head=sprintf("char%sstart%sstop",OFS,OFS,OFS)
  for (h in names) {
    head=sprintf("%s%s%s",head,OFS,h)
  }
  print(head)
  for (l=1; l<i; l++) { # loop over each line/sequence
    line=sprintf("%s%s%s%s%s",seq[l,"chr"],OFS,seq[l,"start"],OFS,seq[l,"end"])
    for (h in names) {
      if (seq[l,h]=="B") line=sprintf("%s%s%s",line,OFS,"B")
      else line=sprintf("%s%s%s",line,OFS,"U")
    }
    if (seq[l,"print"]) print line
  }
}

传递这个命令：

awk -f SO-38563400.awk /c/tmp.txt > /c/Rtable.txt

给予：

$ cat /c/Rtable.txt
char    start   stop    STAT3   ATF3    TEAD4   GATA3   JUND    HNF4A   FOXA2   MAX     CEBPB   SPI1    GABPA   CMYC    P300    E2F1    CTCF    ATF2
chr22   16049850        16050050        B       B       U       B       U       B       B       U       U       U       U       U       B       B       U       B
chr22   16049900        16050100        B       B       B       B       B       B       B       B       B       B       B       B       B       B       B       B

然后读入 r:

> x <- read.table("/c/Rtable.txt", sep="\t",  stringsAsFactors = FALSE, header=T)
> x
char    start     stop STAT3 ATF3 TEAD4 GATA3 JUND HNF4A FOXA2 MAX CEBPB SPI1 GABPA CMYC P300 E2F1 CTCF ATF2
1 chr22 16049850 16050050     B    B     U     B    U     B     B   U     U    U     U    U    B    B    U    B
2 chr22 16049900 16050100     B    B     B     B    B     B     B   B     B    B     B    B    B    B    B    B

请忽略/c/路径的设置，这可以在windows或linux上运行，windows下有awk的端口，由于文件流的操作系统容量，我建议使用linux处理大文件。

我们可以通过在打印结果之前不读取整个文件来节省更多的内存，但这需要一组固定的“名称”，但是你懒得自己提取名称，只是给我发了一堆条目, 练习留给你适应，在 BEGIN 块中制作列表，将其用作每个 seq 的条目，并在每个新的 seq 上打印前一个结果，然后再进行处理。

我希望下次你能花一些时间来提出一个正确的问题，并且你会明白你必须为他人帮助你做出一些努力，特别是在一系列 cmet 要求你改进你的问题之后。

Answer 2

stringr 或 tidyr 可能不是最好的用法，但这可以在 hadleyverse 中以某种易读的方式完成...

逻辑流程是：

• 使用tidyr::fill 和ifelse("Inspecting", rowname, NA) 确定组。
• 将字段更改为您想要的内容
• 使用 reshape (dcast) 获得您想要的格式。

---

library(dplyr)
library(tidyr)
library(reshape2)
library(stringr)

is_in <- function(v1part) {
  return(ifelse(length(v1part) > 0, "B", "U"))
}

ab1<- ab %>% 
  add_rownames() %>%
  mutate(rowname = ifelse(V1=="Inspecting", rowname, NA),
         V4a = ifelse(V4 == "(-)" | V4 == "(+)", NA, V4),

         chr = str_extract_all(ab$V4, "^chr[^:]+", simplify = T)[,1],
         chr = ifelse(chr=="", NA, chr),

         start = str_split_fixed(V4a, ":|-", 3)[,2],
         start = ifelse(start=="", NA, start), 

         stop = str_split_fixed(V4a, ":|-", 3)[,3],
         stop = ifelse(stop=="", NA, stop),

         V1part = str_split_fixed(V1, "\\$|_", 3)[,2]) %>%
  fill(rowname, .direction="down") %>% 
  group_by(rowname) %>%
  fill(chr, .direction="down") %>%
  fill(start, .direction="down") %>%
  fill(stop, .direction="down") %>%
  dcast(chr+start+stop ~ V1part, fun.aggregate=is_in)

> ab1
   chr     start      stop Var.4 ATF3 CEBPB YY1
1 chr1 173244300 173244500     B    B     B   B
2 chr1 173244350 173244550     B    B     B   U

Answer 3

不优雅，但应该可以工作（您的数据有一列带有“|”...我将其命名为 df）：

cond <- which(!df$V2 == "|")
new_df <- data.frame(chr=character(length(cond)), start=character(length(cond)), stop=character(length(cond)))

for (i in 1:length(cond)) {
  line <- df[cond[i], ]
  var <- unlist(strsplit(line$V4, split = ":"))
  var2 <- unlist(strsplit(var[2], split = "-"))
  new_df$chr[i] <- var[1]
  new_df$start[i] <- var2[1]
  new_df$stop[i] <- var2[2]
  for (k in (i+1):(cond[i+1]-1)) {
    # Your code using name <- df$V1 (Use strsplit again)
    # df[i, name] <- ...
  }
}