是否存在R函数可拆分无法完全适配窗口区间的块区间？

用foverlaps合并两个数据框后，存在大量块区间无法完全适配窗口区间的情况（比如块跨多个窗口），需要将这类块拆分为对应窗口内的子区间。例如当窗口为0-100、块为56-139时，拆分为0-100区间内的56-100，以及100-200区间内的100-139。

原始数据 #

window_start_1 window_end_1 block_start_1 block_end_1
 1:              0       100000          5361       70817
 2:              0       100000         71240       94801
 3:              0       100000         95115       95965
 4:             NA           NA         96099      181692
 5:             NA           NA        182089      202832
 6:         200000       300000        203053      203513
 7:         200000       300000        203957      252559
 8:         200000       300000        255798      281602
 9:         200000       300000        282652      285340
10:         200000       300000        285821      286233
11:         200000       300000        286452      286747
12:             NA           NA        287001      307802
13:         300000       400000        308305      318149
14:         300000       400000        319175      323252
15:         300000       400000        323461      333007

期望输出 #

window_start_1 window_end_1 block_start_1 block_end_1
 1:              0       100000          5361       70817
 2:              0       100000         71240       94801
 3:              0       100000         95115       95965
 4:              0       100000         96099      100000
 5:         100000       200000        100000      181692 
 6:         100000       200000        182089      200000
 7:         200000       300000        200000      202832
 8:         200000       300000        203053      203513
 9:         200000       300000        203957      252559
10:         200000       300000        255798      281602
11:         200000       300000        282652      285340
12:         200000       300000        285821      286233
13:         200000       300000        286452      286747
14:         200000       300000        287001      300000
15:         300000       400000        300000      307802
16:         300000       400000        308305      318149
17:         300000       400000        319175      323252
18:         300000       400000        323461      333007

现成工具：IRanges包 #

Bioconductor的IRanges包是专门处理区间数据的工具，能高效完成区间重叠、拆分、交集计算等操作，完全匹配你的需求。

代码示例

# 安装并加载IRanges
if (!requireNamespace("IRanges", quietly = TRUE))
    BiocManager::install("IRanges")
library(IRanges)
library(data.table)

# 原始数据
dt <- data.table(
    window_start_1 = c(0,0,0,NA,NA,200000,200000,200000,200000,200000,200000,NA,300000,300000,300000),
    window_end_1 = c(100000,100000,100000,NA,NA,300000,300000,300000,300000,300000,300000,NA,400000,400000,400000),
    block_start_1 = c(5361,71240,95115,96099,182089,203053,203957,255798,282652,285821,286452,287001,308305,319175,323461),
    block_end_1 = c(70817,94801,95965,181692,202832,203513,252559,281602,285340,286233,286747,307802,318149,323252,333007)
)

# 构建完整的窗口区间（步长100000）
windows <- IRanges(start = seq(0, 300000, by = 100000), end = seq(100000, 400000, by = 100000))
# 转换块数据为IRanges对象
blocks <- IRanges(start = dt$block_start_1, end = dt$block_end_1)

# 找到所有块与窗口的重叠关系
overlaps <- findOverlaps(blocks, windows)

# 计算每个重叠部分的交集（即拆分后的块区间）
intersected_blocks <- pintersect(blocks[queryHits(overlaps)], windows[subjectHits(overlaps)])

# 整理成目标格式
result <- data.table(
    window_start_1 = start(windows[subjectHits(overlaps)]),
    window_end_1 = end(windows[subjectHits(overlaps)]),
    block_start_1 = start(intersected_blocks),
    block_end_1 = end(intersected_blocks)
)

# 按窗口和块起始位置排序
setorder(result, window_start_1, block_start_1)

代码说明

1. IRanges()：将窗口和块数据转换为区间对象，方便后续操作。
2. findOverlaps()：找出所有块与窗口的重叠配对，确定哪些块需要拆分到哪些窗口。
3. pintersect()：计算每个块与对应窗口的交集，得到拆分后的子块区间。
4. 最后将结果转换回data.table格式并排序，得到期望输出。

如果不需要Bioconductor包，也可以用data.table手动实现逻辑，但IRanges是更专业、高效的现成工具。

内容的提问来源于stack exchange，提问作者Nikki