如何用group_by和mutate按分组执行T0减Tx的对数差值计算？

问题背景 #

数据框定义 #

mydf <- structure(list(Time = c("T0", "T3", "T3", "T0", "T3", "T3"), 
    Organism = c("BB", "BB", "BB", "CR", "CR", "CR"), MOF = c("MOF", 
    "MOF", "MOF", "MOF", "MOF", "MOF"), MOFTreatment = c("0", 
    "0", "1", "0", "0", "1"), std_dev_groups = c(" 0.9228677", 
    " 0.8464373", " 0.4491846", " 0.5988814", " 0.5845546", " 1.240182"
    ), Fold_reduction = c(" 1.8958800", " 1.7980552", " 1.3652684", 
    " 1.5145418", " 1.4995760", " 2.362283"), Perc_viability = c("52.7459535", 
    "55.6156459", "73.2456721", "66.0265686", "66.6855175", "42.331930"
    )), class = "data.frame", row.names = c(NA, -6L))

数据展示 #

Time Organism MOF MOFTreatment std_dev_groups Fold_reduction Perc_viability
1   T0       BB MOF            0      0.9228677      1.8958800     52.7459535
2   T3       BB MOF            0      0.8464373      1.7980552     55.6156459
3   T3       BB MOF            1      0.4491846      1.3652684     73.2456721
4   T0       CR MOF            0      0.5988814      1.5145418     66.0265686
5   T3       CR MOF            0      0.5845546      1.4995760     66.6855175
6   T3       CR MOF            1       1.240182       2.362283      42.331930

尝试代码及报错 #

尝试通过分组计算新增Log_loss列，但触发报错：

mydf2 <- mydf %>% group_by(Time, Organism, MOFTreatment) %>%
  mutate(Log_loss = (log10(100/Fold_reduction[Time == "T0"]) - log10(100/Fold_reduction[Time != "T0"])))

报错信息：

Error in `mutate()`:
ℹ In argument: `Log_loss = (...)`.
ℹ In group 1: `Time = "T0"`, `Organism = "BB"`, `MOFTreatment = "0"`.
Caused by error in `100 / Fold_reduction[Time == "T0"]`:
! non-numeric argument to binary operator
Backtrace:
 1. mydf %>% group_by(Time, Organism, MOFTreatment) %>% ...
 3. dplyr:::mutate.data.frame(...)
 4. dplyr:::mutate_cols(.data, dplyr_quosures(...), by)
 6. dplyr:::mutate_col(dots[[i]], data, mask, new_columns)
 7. mask$eval_all_mutate(quo)
 8. dplyr (local) eval()

需要执行的计算逻辑 #

针对每个Organism，用其Time=T0的Fold_reduction值，分别与该Organism下不同MOFTreatment的Time=T3值计算：

# BB的T0对应MOFTreatment=0的T3
log10(100/1.8958800) - log10(100/1.7980552)

# BB的T0对应MOFTreatment=1的T3
log10(100/1.8958800) - log10(100/1.3652684)

# CR的T0对应MOFTreatment=0的T3
log10(100/1.5145418) - log10(100/1.4995760)

# CR的T0对应MOFTreatment=1的T3
log10(100/1.5145418) - log10(100/2.362283)

解决方案 #

错误原因分析 #

1. 数据类型错误：Fold_reduction等列是带空格的字符型，无法直接参与数值运算，导致non-numeric argument报错。
2. 分组逻辑错误：按Time, Organism, MOFTreatment分组后，每个组仅包含单行数据，无法跨Time取数计算。

正确代码 #

library(dplyr)

# 1. 清洗数据：将带空格的字符型数值转为数值型
mydf_clean <- mydf %>%
  mutate(across(c(std_dev_groups, Fold_reduction, Perc_viability), 
                ~ as.numeric(trimws(.))))

# 2. 计算Log_loss列
mydf_final <- mydf_clean %>%
  # 按Organism分组，提取每个生物的T0对应的Fold_reduction值
  group_by(Organism) %>%
  mutate(T0_Fold = Fold_reduction[Time == "T0"]) %>%
  ungroup() %>%
  # 仅对T3行计算Log_loss，T0行设为NA
  mutate(Log_loss = ifelse(Time == "T3",
                           log10(100/T0_Fold) - log10(100/Fold_reduction),
                           NA))

代码说明 #

• 数据清洗：用trimws()去掉数值字符串的空格，再转成数值型，解决运算类型错误。
• 提取基准值：按Organism分组后，提取每个生物Time=T0的Fold_reduction作为基准值T0_Fold。
• 条件计算：仅对Time=T3的行执行目标计算，T0行设为NA，符合需求逻辑。

最终结果示例 #

> mydf_final
# A tibble: 6 × 8
  Time  Organism MOF   MOFTreatment std_dev_groups Fold_reduction Perc_viability Log_loss
  <chr> <chr>    <chr> <chr>                  <dbl>          <dbl>          <dbl>    <dbl>
1 T0    BB       MOF   0                      0.923           1.90           52.7   NA    
2 T3    BB       MOF   0                      0.846           1.80           55.6    0.0217
3 T3    BB       MOF   1                      0.449           1.37           73.2    0.137 
4 T0    CR       MOF   0                      0.599           1.51           66.0   NA    
5 T3    CR       MOF   0                      0.585           1.50           66.7    0.00390
6 T3    CR       MOF   1                      1.24            2.36           42.3   -0.190 

内容的提问来源于stack exchange，提问作者aminards