README.md

In vsbuffalo/gplyr: dplyr helper functions for working with genomic data

Experimental dplyr Extensions for Genomic Range Data

These are some experimental (e.g. beware, things will change... including this uncreative package name) functions that make working with genomic range data held in dataframes (or better, tibbles). If you need more more elaborate genomic range operations in R, you almost certainly want Bioconductor's GenomicRanges.

set.seed(0)
dmel_lens <- tibble::tibble(chrom=c("2L", "X", "3L", "4", "2R", "3R"),
                            length = c(23011544L, 22422827L, 24543557L, 
                                       1351857L, 21146708L, 27905053L))

d <- sim_ranges(20e3, chrom_lengths=dmel_lens)

This object d is a gnibble -- a version of a tibble designed to have the columns chrom, start, end. (Strand is not implemented yet; this could be added later, but if you need strand, use GenomicRanges.)

sim_ranges() automatically appends the chromosome lengths as an attribute to the dataframe. In general, you'll need to do this on your own:

> chrom_lengths(d) <- dmel_lens
> chrom_lengths(d)
# A tibble: 6 × 2
  chrom   length
  <chr>    <int>
1    2L 23011544
2     X 22422827
3    3L 24543557
4     4  1351857
5    2R 21146708
6    3R 27905053

Single-table dplyr verbs should work with gnibbles automatically, passing required metadata in the attributes of this object as needed. gplyr overloads mutate(), as this verb does not. Let's use mutate() to add some random data to each element.

> library(dplyr)
> d <- sim_ranges(20e3, dmel_lens) %>% mutate(x=runif(length(start)))
> class(d)
[1] "gnibble"    "tbl_df"     "tbl"        "data.frame"
> chrom_lengths(d)
# A tibble: 6 × 2
  chrom   length
  <chr>    <int>
1    2L 23011544
2     X 22422827
3    3L 24543557
4     4  1351857
5    2R 21146708
6    3R 27905053

Then, using create_windows(), we can add a column of windows to our dataframe:

> d %>% create_windows(width=10e3)
# A tibble: 20,000 × 5
   chrom  start    end          x          window
   <chr>  <dbl>  <dbl>      <dbl>          <fctr>
1     2L  12054  14424 0.16426985   2L:9996-19991
2     2L  25353  34091 0.51285709  2L:19992-29987
3     2L  28388  31698 0.73325516  2L:19992-29987
4     2L  34465  44224 0.32336798  2L:29988-39984
5     2L  35375  39163 0.32927339  2L:29988-39984
6     2L  47052  51047 0.58059202  2L:39985-49980
7     2L  72249  73636 0.08094737  2L:69974-79969
8     2L  88559  95516 0.96380628  2L:79970-89965
9     2L  99125 102255 0.31359963  2L:89966-99962
10    2L 101196 108285 0.87439077 2L:99963-109958
# ... with 19,990 more rows

With our new window column, we can group the data by chrom, and window, and then summarize

> library(tidyr)
> d %>% create_windows(1e6) %>% 
        group_by(window) %>% 
        summarize_window(xmean=mean(x))
# A tibble: 124 × 3
               window chrom     xmean
               <fctr> <chr>     <dbl>
1         2L:0-958813    2L 0.4964056
2   2L:958814-1917627    2L 0.4932900
3  2L:1917628-2876442    2L 0.5296287
4  2L:2876443-3835256    2L 0.4607771
5  2L:3835257-4794070    2L 0.5052921
6  2L:4794071-5752885    2L 0.5090671
7  2L:5752886-6711699    2L 0.4575886
8  2L:6711700-7670514    2L 0.5106510
9  2L:7670515-8629328    2L 0.4714383
10 2L:8629329-9588142    2L 0.5210464
# ... with 114 more rows

summarize_window() is like dplyr's summarize() except (1) it will group by window if that column is present, and (2) it will carry window variables like chrom forward (and if wstart and wend are present, those too).

Often, we need to refer to a window's start and end positions. We can extract these using separate_window(). Below, we demonstrate this

> dr <- d %>% create_windows(1e6) %>% 
            group_by( window) %>% summarize(xmean=mean(x)) %>%
            separate_window() 
> dr
Source: local data frame [124 x 4]
Groups: chrom [6]

    chrom  wstart    wend     xmean
   <fctr>   <int>   <int>     <dbl>
1      2L       0  958813 0.4964056
2      2L  958814 1917627 0.4932900
3      2L 1917628 2876442 0.5296287
4      2L 2876443 3835256 0.4607771
5      2L 3835257 4794070 0.5052921
6      2L 4794071 5752885 0.5090671
7      2L 5752886 6711699 0.4575886
8      2L 6711700 7670514 0.5106510
9      2L 7670515 8629328 0.4714383
10     2L 8629329 9588142 0.5210464
# ... with 114 more rows
dr <- dr %>% mutate(wcenter=(wstart+wend)/2)
ggplot(dr) + geom_point(aes(x=wcenter, y=xmean)) + facet_wrap(~chrom)

This workflow is also encapsulated in the convenience function, `append_wcenter():

> d %>% create_windows(1e6) %>% 
        group_by(window) %>% 
        summarize_window(xmean=mean(x)) %>%
        append_wcenter()
# A tibble: 124 × 6
    chrom  wstart    wend             window     xmean   wcenter
   <fctr>   <int>   <int>             <fctr>     <dbl>     <dbl>
1      2L       0  958813        2L:0-958813 0.4964056  479406.5
2      2L  958814 1917627  2L:958814-1917627 0.4932900 1438220.5
3      2L 1917628 2876442 2L:1917628-2876442 0.5296287 2397035.0
4      2L 2876443 3835256 2L:2876443-3835256 0.4607771 3355849.5
5      2L 3835257 4794070 2L:3835257-4794070 0.5052921 4314663.5
6      2L 4794071 5752885 2L:4794071-5752885 0.5090671 5273478.0
7      2L 5752886 6711699 2L:5752886-6711699 0.4575886 6232292.5
8      2L 6711700 7670514 2L:6711700-7670514 0.5106510 7191107.0
9      2L 7670515 8629328 2L:7670515-8629328 0.4714383 8149921.5
10     2L 8629329 9588142 2L:8629329-9588142 0.5210464 9108735.5
# ... with 114 more rows

Or, if you want all chromosomes concatenated on a single x-axis, use append_wcumpos().

dr <- d %>% create_windows(1e6) %>% arrange(chrom) %>% 
            group_by(chrom, window) %>% summarize(xmean=mean(x)) %>%
            ungroup() %>%
            separate_window() %>% 
            append_wcumpos()
ggplot(dr) + geom_point(aes(x=wcumpos, y=xmean, color=chrom)) 

vsbuffalo/gplyr documentation built on May 3, 2019, 7:07 p.m.