Question

我正在使用R中的dplyr包来过滤我的基因表达数据。我已经计算出倍数变化，并希望过滤至少一个样本（列）的值大于+0.584963或小于-0.584963的基因（行）。示例数据：

       X SAMPLE_1_FC SAMPLE_2_FC SAMPLE_3_FC SAMPLE_4_FC SAMPLE_5_FC
GENE_1      0.6780      0.4050      0.8870      0.3300      0.2230
GENE_2      0.2340     -0.6670      0.0020      0.1240      0.3560
GENE_3      0.0170      0.1560      0.1120      0.0080     -0.1230
GENE_4     -0.0944     -0.1372     -0.1800     -0.2228     -0.2656
GENE_5     -0.8080     -0.7800     -0.5560      0.0340      0.4450
GENE_6      0.2091      0.1106      0.0121     -0.0864     -0.1849
GENE_7      0.5980      0.7680      0.9970      0.4670     -0.7760

我当前正在使用以下脚本

det.cols<- colnames(my.data)[which(grepl("fc",tolower(colnames(my.data))))]
filt <- gsub(","," | ",toString(paste("`",det.cols,"`",">abs(0.584963)", sep = "")))
my.datasub<- my.data %>% filter_(filt)

但是这仅返回大于+0.584963的基因，而不返回负基因。在该示例的情况下，我想要的是具有基因1、2、5和7的子集列表。但是相反，它仅给我基因1和7。我该如何更改？

我希望答案采用以下格式：

 X SAMPLE_1_FC SAMPLE_2_FC SAMPLE_3_FC SAMPLE_4_FC SAMPLE_5_FC
GENE_1      0.6780      0.4050      0.8870      0.3300      0.2230
GENE_2      0.2340     -0.6670      0.0020      0.1240      0.3560
GENE_5     -0.8080     -0.7800     -0.5560      0.0340      0.4450
GENE_7      0.5980      0.7680      0.9970      0.4670     -0.7760

谢谢。

Answer 1

使用filter_at中的dplyr可能是一种更加灵活的方法...

# set up sample data with 50000 rows [as proposed by Arthur Yip above]
mydata <- tibble(X = c("GENE_1", "GENE_2", "GENE_3", "GENE_4", "GENE_5", "GENE_6", "GENE_7", 1:50000),
                     SAMPLE_1_FC = c(0.678, 0.234, 0.017, -0.0944, -0.808, 0.2091, 0.598, rnorm(50000, 0, 1)),
                     SAMPLE_2_FC = c(0.405, -0.667, 0.156, -0.1372, -0.78, 0.1106, 0.768, rnorm(50000, 0, 1)),
                     SAMPLE_3_FC = c(0.887, 0.002, 0.112, -0.18, -0.556, 0.0121, 0.997, rnorm(50000, 0, 1)),
                     SAMPLE_4_FC = c(0.33, 0.124, 0.008, -0.2228, 0.034, -0.0864, 0.467, rnorm(50000, 0, 1)),
                     SAMPLE_5_FC = c(0.223, 0.356, -0.123, -0.2656, 0.445, -0.1849, -0.776, rnorm(50000, 0, 1)))

# duplicate 30 more columns [as proposed by Arthur Yip above]
mydata2 <- bind_cols(mydata, mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6])

mydata2 %>%
  filter_at(vars(contains("fc")), .vars_predicate =  any_vars(abs(.) > 0.584963))

在vars()中，您可以定义要对其应用过滤的变量列表。在.vars_predicate之后，您可以定义过滤条件（any_vars等于|，all_vars等于&）。

Answer 2

长话短说，您在代码中的错误位置放置了abs()。

我在这里固定了它

det.cols<- colnames(my.data)[which(grepl("fc",tolower(colnames(my.data))))]
filt <- gsub(","," | ",toString(paste("abs(`",det.cols,"`)",">0.584963", sep = "")))
my.datasub<- my.data %>% filter_(filt)

为了获得更大的灵活性，@ ha_pu提供了一个很棒的filter_at解决方案，该解决方案以我之前的解决方案为基础（在我确定您的代码中的错误之前）。

Answer 3

这是一个可以灵活处理样本和数据行数的解决方案。它涉及将数据转换为长格式，然后过滤基因和特定样本。我在50k个基因和35个样本上对其进行了测试，并且运行了<1秒。

library(tidyverse)

# set up sample data with 50000 rows
mydata <- data.frame(stringsAsFactors=FALSE,
                     X = c("GENE_1", "GENE_2", "GENE_3", "GENE_4", "GENE_5", "GENE_6", "GENE_7", 1:50000),
                     SAMPLE_1_FC = c(0.678, 0.234, 0.017, -0.0944, -0.808, 0.2091, 0.598, rnorm(50000, 0, 1)),
                     SAMPLE_2_FC = c(0.405, -0.667, 0.156, -0.1372, -0.78, 0.1106, 0.768, rnorm(50000, 0, 1)),
                     SAMPLE_3_FC = c(0.887, 0.002, 0.112, -0.18, -0.556, 0.0121, 0.997, rnorm(50000, 0, 1)),
                     SAMPLE_4_FC = c(0.33, 0.124, 0.008, -0.2228, 0.034, -0.0864, 0.467, rnorm(50000, 0, 1)),
                     SAMPLE_5_FC = c(0.223, 0.356, -0.123, -0.2656, 0.445, -0.1849, -0.776, rnorm(50000, 0, 1)))

# duplicate 30 more columns
mydata2 <- bind_cols(mydata, mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6], mydata[2:6])

(mydata3 <- mydata2 %>% gather(key = "sample_num", value = "fc", 2:length(mydata)) %>%
  filter(fc > 0.584963 | fc < -0.584963) %>%
  select(X) %>%
  arrange(desc(X)) %>%
  unique() %>%
  head())
#>         X
#> 1  GENE_7
#> 5  GENE_5
#> 7  GENE_2
#> 8  GENE_1
#> 10   9999
#> 14   9998

^{由reprex package（v0.2.1）于2019-03-01创建}

R dplyr过滤值大于+ N且小于-N的数据：abs（）函数？

3 个答案: