匹配行，在 R 中按组创建排名和求和排名

Question

我有一个包含约 30,000 行和约 17,000 列的庞大数据集，以及一个包含 character 元素的向量。

这是一个重新创建我的数据集的虚拟集

### Example

df <- data.frame(Gene=paste0("gene", 1:60), replicate(60, runif(60, min=0, max=100)))
colnames(df) <- c("GeneName", paste0("TisA.", 1:20), paste0("TisB.", 1:20), paste0("TisC.", 1:20))

genes <- sample(df$GeneName, 5)

head(df)
#      GeneName    TisA.1    TisA.2    TisA.3   TisA.4
#1    gene1  1.987621 17.936562 18.145417 59.43023
#2    gene2 60.031713 73.822846 93.946769 72.27633
#3    gene3 44.833748 47.890719 77.100497 39.45719
#4    gene4 44.662776 26.285659 30.087606 49.50682
#5    gene5 63.770411  6.469006  3.797708 68.17532

我需要为数据框匹配向量中的元素，这很容易完成

 df.new <- df[df$GeneName %in% genes,]

然后，我想要的是，对于每个genes，为每个基因创建等级值，然后将等级与Tis (A, B, C) 相加

例如，我可以使用 gene 对值进行排序

genes.ord <- sort(df.new[1,], decreasing = TRUE)

但是，我被困在这里，这将是为基因分配等级并按组求和这些等级的最快方法，即TisA、TisB 和TisC？

为澄清起见，每组有 20 个样本TisA.1, TisA.2, ..., TisA.20

期望的输出是：

 GeneName   TisA TisB TisC
    gene4     24   32   10 ## these are random values to show sum of ranks for each of genes in the vector
    gene1     14   12   20 ## these are random values to show sum of ranks for each of genes in the vector
   gene40      4   92   12 ## these are random values to show sum of ranks for each of genes in the vector
    gene2     64    2   40 ## these are random values to show sum of ranks for each of genes in the vector
   gene15     84   32    9 ## these are random values to show sum of ranks for each of genes in the vector

P.S 我的真实数据集中的一些值可以是 0 并且在不同的列中重复

Answer 1

直接使用 tidyverse

# your data. Including seed to make it reproducible
set.seed(123)
df <- data.frame(Gene=paste0("gene", 1:60), replicate(60, runif(60, min=0, max=100)))
colnames(df) <- c("GeneName", paste0("TisA.", 1:20), paste0("TisB.", 1:20), paste0("TisC.", 1:20))

library(tidyverse)
as.tbl(df) %>% 
    gather(key, value, -GeneName) %>% 
    group_by(GeneName) %>% 
    mutate(Ranks = rank(value, ties.method = "first"))  %>% 
    separate(key, into = c("key1", "key2"), sep = "[.]") %>% 
    group_by(GeneName,key1) %>% 
    summarise(Sum=sum(Ranks)) %>% 
    spread(key1, Sum)
# A tibble: 60 x 4
# Groups:   GeneName [60]
GeneName  TisA  TisB  TisC
*   <fctr> <int> <int> <int>
1    gene1   698   620   512
2   gene10   525   653   652
3   gene11   631   598   601
4   gene12   487   679   664
5   gene13   688   579   563
6   gene14   674   581   575
7   gene15   618   647   565
8   gene16   696   552   582
9   gene17   656   560   614
10  gene18   543   649   638 

或者试试 baseR 解决方案...有点复杂

df1 <- apply(df[-1], 1, rank, ties.method= "first")
df2 <- apply(df1, 2, function(x){
  aggregate(x, list(sapply(strsplit(colnames(df), "[.]"), "[", 1)[-1]), sum)
  })
df3 <- cbind.data.frame(df$GeneName, t(Reduce(cbind, lapply(df2, "[", 2))))
colnames(df3) <- c("GeneName",  "TisA", "TisB", "TisC")
head(df3[order(df3$GeneName),])
GeneName TisA TisB TisC
   gene1  698  620  512
  gene10  525  653  652
  gene11  631  598  601
  gene12  487  679  664
  gene13  688  579  563
  gene14  674  581  575