如何更高效编写str_detect代码？精简数据提取适配UpSet图

需要从数据框的某列提取指定字符对应的行名，每个匹配字符对应的行名存入单独的对象，用于绘制UpSet图展示各集合的重叠情况。现有代码冗长重复，需精简优化。

模拟数据（修正后可运行） #

# 修正rep用法，生成100行有效数据
name <- rep(c("rs123", "rs124", "rs125", "rs126", "rs127", "rs128", "rs129", "rs130"), length.out = 100)
source <- rep("dbSNP", 100)
chromosome <- rep("17", 100)
evidence <- rep(c("TOPMed", "Frequency", "Cited", "Frequency, Cited", "Disease", "ESP", "TOPMed", "GWAS", "1000Genome"), length.out = 100)

Biomart <- data.frame(name, source, chromosome, evidence)

现有冗余代码 #

'1000Genome' <- Biomart[str_detect(Biomart$evidence,"1000Genomes"),]$name 
'freq' <- Biomart[str_detect(Biomart$evidence,"Frequency"),]$name
'TOPMed' <-  Biomart[str_detect(Biomart$evidence,"TOPMed"),]$name
'gnomAD' <- Biomart[str_detect(Biomart$evidence, "gnomAD"),]$name
'Cited' <- Biomart[str_detect(Biomart$evidence, "Cited"),]$name
'ESP' <- Biomart[str_detect(Biomart$evidence, "ESP"),]$name
'ExAC' <- Biomart[str_detect(Biomart$evidence, "ExAC"),]$name
'Phenotype_or_Disease' <- Biomart[str_detect(Biomart$evidence, "Disease"),]$name

精简方案 #

方案1：用循环生成集合列表 #

把需要匹配的字符存入向量，通过循环批量生成每个集合的行名，结果存入列表（方便统一管理和调用）：

library(stringr)

# 定义需要匹配的证据类型：键为集合名，值为匹配字符串
evidence_types <- list(
  "1000Genome" = "1000Genome",
  "freq" = "Frequency",
  "TOPMed" = "TOPMed",
  "gnomAD" = "gnomAD",
  "Cited" = "Cited",
  "ESP" = "ESP",
  "ExAC" = "ExAC",
  "Phenotype_or_Disease" = "Disease"
)

# 循环生成每个集合的行名，存入列表
evidence_sets <- lapply(evidence_types, function(pattern) {
  Biomart[str_detect(Biomart$evidence, pattern), "name"]
})

# 调用单个集合示例：evidence_sets$TOPMed

方案2：生成UpSet图直接可用的二进制矩阵 #

UpSet图通常需要每行对应一个样本，每列对应一个集合，值为1/0表示样本是否属于该集合，这种格式可直接传入UpSetR或ComplexUpset包，省去后续整理步骤：

library(stringr)
library(dplyr)

# 定义集合名称和对应匹配字符串
evidence_cols <- c(
  "1000Genome", "freq", "TOPMed", "gnomAD", 
  "Cited", "ESP", "ExAC", "Phenotype_or_Disease"
)
patterns <- c(
  "1000Genome", "Frequency", "TOPMed", "gnomAD", 
  "Cited", "ESP", "ExAC", "Disease"
)

# 生成二进制矩阵
upset_data <- Biomart %>%
  select(name) %>%
  mutate(across(all_of(evidence_cols), ~str_detect(Biomart$evidence, patterns[which(evidence_cols == cur_column())]))) %>%
  mutate(across(-name, as.integer))

# 查看结果示例
head(upset_data)

说明 #

• 方案1保留了「每个集合单独存储」的需求，用列表替代多个独立变量，新增集合只需修改evidence_types，维护性更强。
• 方案2直接生成UpSet图所需格式，一步到位，是更高效的选择。

内容的提问来源于stack exchange，提问作者Ted