R语言如何按指定列条件删除数据框中的唯一峰交集行

问题描述 #

我有一个包含染色体坐标、峰交集及其他相关信息的数据框，总计127471行、11列，缩略表如下：

需要删除所有属于唯一峰交集的行，判定规则为：同一Union区间（UnionChr、UnionStart、UnionEnd三个坐标完全一致的区间）中只要存在Overlap值为0的行，该区间下所有行都需要删除。示例中第1-2行、8-9行、10-11行所属区间均包含Overlap=0的记录，因此全部删除，仅保留第3-7行。

此前尝试过子集筛选、for循环、duplicated()、unique()等方法均未实现目标，因不熟悉Python，仅需单次完成处理，故寻求R语言可行方案。

示例数据构造代码 #

df <- data.frame(UnionChr=c("chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1"), 
                  UnionStart = c(3657227, 3657227, 3658443, 3658443, 3658443, 3665636, 3665636, 4468858, 4468858, 4472410, 4472410), 
                  UnionEnd = c(3658092, 3658092, 3664519, 3664519, 3664519, 3666032, 3666032, 4469245, 4469245, 4473380, 4473380), 
                  IntersectChr = c("." , "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", "chr1", ".",  ".", "chr1"), 
                  IntersectStart = c(-1, 3657227, 3658443, 3663340, 3658833, 3665705, 3665636, 4468858, -1 , -1, 4472410), 
                  IntersectEnd = c( -1, 3658092, 3662838, 3664519, 3664156, 3666032, 3665919, 4469245, -1, -1, 4473380), 
                  IntersectName = c(".", "dko_k27_peak_1", "wt_k27_peak_1", "wt_k27_peak_2",  "dko_k27_peak_2", "wt_k27_peak_3", "dko_k27_peak_3", "wt_k27_peak_4",  ".", ".", "dko_k27_peak_4"), 
                  Overlap = c(0, 865, 4395, 1179, 5323, 327, 283, 387, 0, 0, 970), 
                  Genotype = c("WT", "DKO", "WT", "WT", "DKO", "WT", "DKO", "WT", "DKO", "WT", "DKO"))

实现方案 #

不需要写复杂循环，按Union区间分组过滤即可，两种实现方式都可以高效处理12万行数据，运行时间不到1秒：

• 基础R实现（无需安装任何第三方包，适合单次快速处理）

# 为每个Union区间生成唯一分组ID
group_id <- paste(df$UnionChr, df$UnionStart, df$UnionEnd, sep = "_")
# 提取所有包含Overlap=0的无效区间
invalid_groups <- unique(group_id[df$Overlap == 0])
# 筛选保留非无效区间的所有行
df_filtered <- df[!group_id %in% invalid_groups, ]

• dplyr实现（适合习惯用tidyverse语法的场景）

# 首次使用需先安装包
# install.packages("dplyr")
library(dplyr)

df_filtered <- df %>%
  group_by(UnionChr, UnionStart, UnionEnd) %>%
  filter(!any(Overlap == 0)) %>%
  ungroup()

运行以上代码后得到的df_filtered和预期结果完全一致，示例数据中会保留第3-7行共5条有效记录。

内容的提问来源于stack exchange，提问作者ispeakcat