缘由
统计是生信的基石之一
碎碎念
要爱一个具体的人，不要爱一个抽象的人； 要爱生活，不要爱生活的意义。 ——《卡拉马佐夫兄弟》
Ref
GGPUBR: How to Add Adjusted P-values to a Multi-Panel GGPlot - Datanovia
基因表达差异分析完整教程
1. 数据加载与预处理
加载必要的R包
library(ggpubr) # 数据可视化
library(rstatix) # 统计检验
library(tidyverse) # 数据处理创建模拟数据集
# Data preparation
df <- tibble::tribble(
~sample_type, ~expression, ~cancer_type, ~gene,
"cancer", 25.8, "Lung", "Gene1",
"cancer", 25.5, "Liver", "Gene1",
"cancer", 22.4, "Liver", "Gene1",
"cancer", 21.2, "Lung", "Gene1",
"cancer", 24.5, "Liver", "Gene1",
"cancer", 27.3, "Liver", "Gene1",
"cancer", 30.9, "Liver", "Gene1",
"cancer", 17.6, "Breast", "Gene1",
"cancer", 19.7, "Lung", "Gene1",
"cancer", 9.7, "Breast", "Gene1",
"cancer", 15.2, "Breast", "Gene2",
"cancer", 26.4, "Liver", "Gene2",
"cancer", 25.8, "Lung", "Gene2",
"cancer", 9.7, "Breast", "Gene2",
"cancer", 21.2, "Lung", "Gene2",
"cancer", 24.5, "Liver", "Gene2",
"cancer", 14.5, "Breast", "Gene2",
"cancer", 19.7, "Lung", "Gene2",
"cancer", 25.2, "Lung", "Gene2",
"normal", 43.5, "Lung", "Gene1",
"normal", 76.5, "Liver", "Gene1",
"normal", 21.9, "Breast", "Gene1",
"normal", 69.9, "Liver", "Gene1",
"normal", 101.7, "Liver", "Gene1",
"normal", 80.1, "Liver", "Gene1",
"normal", 19.2, "Breast", "Gene1",
"normal", 49.5, "Lung", "Gene1",
"normal", 34.5, "Breast", "Gene1",
"normal", 51.9, "Lung", "Gene1",
"normal", 67.5, "Lung", "Gene2",
"normal", 30, "Breast", "Gene2",
"normal", 76.5, "Liver", "Gene2",
"normal", 88.5, "Liver", "Gene2",
"normal", 69.9, "Liver", "Gene2",
"normal", 49.5, "Lung", "Gene2",
"normal", 80.1, "Liver", "Gene2",
"normal", 79.2, "Liver", "Gene2",
"normal", 12.6, "Breast", "Gene2",
"normal", 97.5, "Liver", "Gene2",
"normal", 64.5, "Liver", "Gene2"
)
- 包含4个变量：
sample_type: 样本类型（癌症/正常）expression: 基因表达量cancer_type: 癌症类型gene: 基因名称
2. 描述性统计分析
# Summary statistics
df %>%
group_by(gene, cancer_type, sample_type) %>%
get_summary_stats(expression, type = "common")group_by(): 按基因、癌症类型和样本类型分组get_summary_stats(): 计算常用统计量- 输出包括：样本数、均值、标准差、标准误、中位数等
- 帮助了解数据的整体分布情况
- 为后续统计检验提供基础信息
3. 统计检验与多重检验校正
执行t检验
stat.test <- df %>%
group_by(cancer_type, gene) %>% # 按癌症类型和基因分组
t_test(expression ~ sample_type) %>% # 执行t检验
adjust_pvalue(method = "bonferroni") %>% # P值校正
add_significance() # 添加显著性标记
stat.test-
t_test()函数：
- 公式
expression ~ sample_type表示比较不同样本类型的表达差异 - 自动执行两组间的t检验
- 公式
-
多重检验校正：
adjust_pvalue(method = "bonferroni"): 使用Bonferroni法校正- 为什么需要校正？ 当进行多次检验时，假阳性率会增加
- 其他可选方法：BH（FDR）、holm、hochberg等
-
显著性标记：
add_significance(): 自动添加星号标记*p < 0.05,**p < 0.01,***p < 0.001
4. 数据可视化
创建基础箱线图
bxp <- ggboxplot(
df, x = "sample_type", y = "expression",
facet.by = c("gene", "cancer_type") # 按基因和癌症类型分面
) +
rotate_x_text(angle = 60) # x轴文字旋转60度添加P值位置信息
stat.test <- stat.test %>% add_xy_position(x = "sample_type")多种P值显示方式
方式1：显示具体P值
bxp + stat_pvalue_manual(stat.test, label = "p.adj")适用场景：需要精确P值的学术论文
方式2：显示显著性标记
bxp + stat_pvalue_manual(stat.test, label = "p.adj.signif")适用场景：快速识别显著差异
方式3：隐藏不显著结果
bxp + stat_pvalue_manual(
stat.test, label = "p.adj.signif",
hide.ns = TRUE, tip.length = 0
)参数说明：
hide.ns = TRUE: 隐藏不显著的结果tip.length = 0: 设置指示线尖端长度为0
方式4：组合显示P值和显著性
bxp + stat_pvalue_manual(
stat.test, label = "{p.adj}{p.adj.signif}",
hide.ns = TRUE, tip.length = 0
)显示效果：同时显示具体P值和星号标记