CODEWITHSUNDEEP
rdata.tablebioinformaticsoverlap
DN1
DN1

Reputation: 218

How to filter rows by column value ranges in R?

I have 2 genetic datasets. One that defines ranges in the genome per row, and another dataset that is rows of gene length ranges that I want to make sure do not have any overlap with the ranges in the first dataset.

For example, my data looks like this:

#df1:
Chromosome     Min      Max
1              10       500 
1              450      550
2              20       100
2              900      1500
2              200      210
3               5       15
4              10       20

#df2:
Gene   Gene.Start    Gene.End   Chromosome
Gene1   10             60           1
Gene2   950            990          1
Gene3   8              14           3

I want to pull out/select rows in df2 that do not have a Gene.Start and Gene.End range where anything in the range falls in the ranges given in df1 in the Min and Max columns - with, importantly, the consideration of the Chromosome number must also match.

The expected output from the example would look like:

Gene   Gene.Start    Gene.End   Chromosome
Gene2   950            990          1

Gene2 is the only gene/row with a start and end range that doesn't fall in any ranges with matching Chromosome (looking at ranges in Chromosome 1) in df1.

To code this I am trying with data.table but I'm not sure how to get the ranges to be considered like I want them to.

I've been trying to get this working but I'm not sure what I'm doing:

df2[df1, match := i.Gene,
                 on = .(Chromosome, (df2$Gene.Start > & < df2$Gene.End) > Min, (df2$Gene.Start > & < df2$Gene.End) < Max)]

Error: unexpected '&'

What can I do to filter a dataframe by its ranges depending on ranges in another dataframe?

Example input data:

df1 <- structure(list(Chromosome = c(1L, 1L, 2L, 2L, 2L, 3L, 4L), Min = c(10L, 
450L, 20L, 900L, 200L, 5L, 10L), Max = c(500L, 550L, 100L, 1500L, 
210L, 15L, 20L)), row.names = c(NA, -7L), class = c("data.table", 
"data.frame"))

df2 <- structure(list(Gene = c("Gene1", "Gene2", "Gene3"), Gene.Start = c(10L, 
950L, 8L), Gene.End = c(60L, 990L, 14L), Chromosome = c(1L, 1L, 
3L)), row.names = c(NA, -3L), class = c("data.table", "data.frame"
))

Upvotes: 2

Views: 675

Answers (3)

DN1
DN1

Reputation: 218

The data.table solution works best as it's the fastest on my much larger real data, but I did end up finding another solution with GenomicRanges so I thought I'd also share for anyone else's future reference:

library(GenomicRanges)

gr1 <- makeGRangesFromDataFrame(
  data.frame(
    chr=df1$Chromosome,
    start=df1$Min,
    end=df1$Max),
  keep.extra.columns=TRUE)

gr2 <- makeGRangesFromDataFrame(
  data.frame(
    chr=df2$Chromosome,
    start=df2$Gene.Start,
    end=df2$Gene.End,
    Gene = df2$Gene),
  keep.extra.columns=TRUE)

no_overlaps <- gr2[-queryHits(findOverlaps(gr2, gr1, type="any")),] 

no_overlap_genes <- unique(no_overlaps$Gene)

gene_matches <- df2[Gene %in% no_overlap_genes]

Upvotes: 0

TarJae
TarJae

Reputation: 78917

Here is my try with dplyr approach. Please let me know.

library(dplyr)
library(tidyr)
df2 %>% 
  right_join(df1, by = "Chromosome") %>% 
  filter(Gene.Start<Min | Gene.Start>Max, Gene.End>Max | Gene.End>Min) %>% 
  distinct(Gene, Gene.Start, Gene.End, Chromosome, .keep_all = TRUE) %>% 
  select(Gene, Gene.Start, Gene.End, Chromosome)

Output:

   Gene Gene.Start Gene.End Chromosome
1 Gene2        950      990          1

Upvotes: 2

Wimpel
Wimpel

Reputation: 27732

Here is a data.table approach

library(data.table)
# keep Gene that are not joined in the non-equi join on df1 below
df2[!Gene %in% df2[df1, on = .(Chromosome, Gene.Start >= Min, Gene.End <= Max)]$Gene, ]
#     Gene Gene.Start Gene.End Chromosome
# 1: Gene2        950      990          1

Upvotes: 3

Related Questions