如何在R中对数据框每行的人类组与小鼠组列执行t检验

针对你的需求，这里提供几种高效的R语言实现方法，适用于大型数据集：

方法一：基础R原生实现 #

适合习惯基础R语法的用户，无需额外安装包：

# 将基因名设为行名，方便后续匹配
rownames(df) <- df$genes
# 提取数值列（排除基因名列）
df_numeric <- df[, -1]

# 定义针对单行基因的t检验函数
run_t_test <- function(row_data) {
  # 拆分人类组和小鼠组的数值
  human_group <- c(row_data["human_A"], row_data["human_B"], row_data["human_C"])
  mouse_group <- c(row_data["mouse_A"], row_data["mouse_B"])
  
  # 执行t检验（默认Welch检验，若需方差齐性检验加var.equal=TRUE）
  test_result <- t.test(human_group, mouse_group)
  
  # 返回关键统计结果
  data.frame(
    gene = rownames(df_numeric)[which(df_numeric == row_data, arr.ind = TRUE)[1, 1]],
    t_value = round(test_result$statistic, 4),
    p_value = round(test_result$p.value, 6),
    human_mean = round(mean(human_group), 3),
    mouse_mean = round(mean(mouse_group), 3),
    mean_difference = round(mean(human_group) - mean(mouse_group), 3)
  )
}

# 对每行基因应用函数，合并结果
final_results <- do.call(rbind, apply(df_numeric, 1, run_t_test))

# 查看前6行结果
head(final_results)

方法二：tidyverse管道式实现 #

代码更直观，适合数据清洗与分析一体化流程：

library(tidyverse)

# 将宽格式数据转换为长格式，添加分组标签
df_long <- df %>%
  pivot_longer(cols = -genes, names_to = "sample", values_to = "count") %>%
  mutate(group = ifelse(str_detect(sample, "human"), "human", "mouse"))

# 按基因分组执行t检验，提取统计量
t_test_results <- df_long %>%
  group_by(genes) %>%
  summarise(
    t_test_obj = list(t.test(count ~ group)),
    t_value = round(t_test_obj[[1]]$statistic, 4),
    p_value = round(t_test_obj[[1]]$p.value, 6),
    human_mean = round(mean(count[group == "human"]), 3),
    mouse_mean = round(mean(count[group == "mouse"]), 3),
    mean_difference = round(human_mean - mouse_mean, 3)
  ) %>%
  select(-t_test_obj) # 移除临时的检验对象列

# 查看结果
head(t_test_results)

方法三：data.table高效实现 #

处理超大型数据集（如十万级基因）时速度最优：

library(data.table)

# 转换为data.table格式
setDT(df)
# 转长格式并添加分组
df_long <- melt(df, id.vars = "genes", variable.name = "sample", value.name = "count")
df_long[, group := fifelse(grepl("human", sample), "human", "mouse")]

# 按基因分组执行t检验
t_test_results <- df_long[, {
  test <- t.test(count ~ group)
  .(
    t_value = round(test$statistic, 4),
    p_value = round(test$p.value, 6),
    human_mean = round(mean(count[group == "human"]), 3),
    mouse_mean = round(mean(count[group == "mouse"]), 3),
    mean_difference = round(human_mean - mouse_mean, 3)
  )
}, by = genes]

# 查看结果
head(t_test_results)

关键补充说明 #

• 方差齐性与检验类型：默认使用Welch t检验（不假设方差齐），若确定两组方差齐，添加var.equal = TRUE参数切换为Student's t检验。
• 非参数替代：若数据不满足正态分布，将t.test替换为wilcox.test即可执行 Wilcoxon 秩和检验。
• 多重检验校正：由于同时检验多个基因，建议对p值做校正（如FDR校正），示例代码：

  # 以tidyverse版本为例添加校正p值
  t_test_results <- t_test_results %>%
    mutate(adj_p_value = round(p.adjust(p_value, method = "fdr"), 6))
  

内容的提问来源于stack exchange，提问作者LDT