/plyranges

A grammar of genomic data transformation

Primary LanguageR

plyranges: fluent genomic data analysis

R-CMD-check-bioc BioC status Codecov test coverage

plyranges provides a consistent interface for importing and wrangling genomics data from a variety of sources. The package defines a grammar of genomic data transformation based on dplyr and the Bioconductor packages IRanges, GenomicRanges, and rtracklayer. It does this by providing a set of verbs for developing analysis pipelines based on Ranges objects that represent genomic regions:

  • Modify genomic regions with the mutate() and stretch() functions.
  • Modify genomic regions while fixing the start/end/center coordinates with the anchor_ family of functions.
  • Sort genomic ranges with arrange().
  • Modify, subset, and aggregate genomic data with the mutate(), filter(), and summarise()functions.
  • Any of the above operations can be performed on partitions of the data with group_by().
  • Find nearest neighbour genomic regions with the join_nearest_ family of functions.
  • Find overlaps between ranges with the join_overlaps_ family of functions.
  • Merge all overlapping and adjacent genomic regions with reduce_ranges().
  • Merge the end points of all genomic regions with disjoin_ranges().
  • Import and write common genomic data formats with the read_/write_ family of functions.

For more details on the features of plyranges, read the vignette. For a complete case-study on using plyranges to combine ATAC-seq and RNA-seq results read the fluentGenomics workflow.

Installation

plyranges can be installed from the latest Bioconductor release:

# install.packages("BiocManager")
BiocManager::install("plyranges")

To install the development version from GitHub:

BiocManager::install("sa-lee/plyranges")

Quick overview

About Ranges

Ranges objects can either represent sets of integers as IRanges (which have start, end and width attributes) or represent genomic intervals (which have additional attributes, sequence name, and strand) as GRanges. In addition, both types of Ranges can store information about their intervals as metadata columns (for example GC content over a genomic interval).

Ranges objects follow the tidy data principle: each row of a Ranges object corresponds to an interval, while each column will represent a variable about that interval, and generally each object will represent a single unit of observation (like gene annotations).

We can construct a IRanges object from a data.frame with a start or width using the as_iranges() method.

library(plyranges)
df <- data.frame(start = 1:5, width = 5)
as_iranges(df)
#> IRanges object with 5 ranges and 0 metadata columns:
#>           start       end     width
#>       <integer> <integer> <integer>
#>   [1]         1         5         5
#>   [2]         2         6         5
#>   [3]         3         7         5
#>   [4]         4         8         5
#>   [5]         5         9         5
# alternatively with end
df <- data.frame(start = 1:5, end = 5:9)
as_iranges(df)
#> IRanges object with 5 ranges and 0 metadata columns:
#>           start       end     width
#>       <integer> <integer> <integer>
#>   [1]         1         5         5
#>   [2]         2         6         5
#>   [3]         3         7         5
#>   [4]         4         8         5
#>   [5]         5         9         5

We can also construct a GRanges object in a similar manner. Note that a GRanges object requires at least a seqnames column to be present in the data.frame (but not necessarily a strand column).

df <- data.frame(seqnames = c("chr1", "chr2", "chr2", "chr1", "chr2"),
                 start = 1:5,
                 width = 5)
as_granges(df)
#> GRanges object with 5 ranges and 0 metadata columns:
#>       seqnames    ranges strand
#>          <Rle> <IRanges>  <Rle>
#>   [1]     chr1       1-5      *
#>   [2]     chr2       2-6      *
#>   [3]     chr2       3-7      *
#>   [4]     chr1       4-8      *
#>   [5]     chr2       5-9      *
#>   -------
#>   seqinfo: 2 sequences from an unspecified genome; no seqlengths
# strand can be specified with `+`, `*` (mising) and `-`
df$strand <- c("+", "+", "-", "-", "*")
as_granges(df)
#> GRanges object with 5 ranges and 0 metadata columns:
#>       seqnames    ranges strand
#>          <Rle> <IRanges>  <Rle>
#>   [1]     chr1       1-5      +
#>   [2]     chr2       2-6      +
#>   [3]     chr2       3-7      -
#>   [4]     chr1       4-8      -
#>   [5]     chr2       5-9      *
#>   -------
#>   seqinfo: 2 sequences from an unspecified genome; no seqlengths

Example: finding GWAS hits that overlap known exons

Let’s look at a more a realistic example (taken from HelloRanges vignette).

Suppose we have two GRanges objects: one containing coordinates of known exons and another containing SNPs from a GWAS.

The first and last 5 exons are printed below, there are two additional columns corresponding to the exon name, and a score.

We could check the number of exons per chromosome using group_by and summarise.

exons
#> GRanges object with 459752 ranges and 2 metadata columns:
#>            seqnames            ranges strand |
#>               <Rle>         <IRanges>  <Rle> |
#>        [1]     chr1       11874-12227      + |
#>        [2]     chr1       12613-12721      + |
#>        [3]     chr1       13221-14409      + |
#>        [4]     chr1       14362-14829      - |
#>        [5]     chr1       14970-15038      - |
#>        ...      ...               ...    ... .
#>   [459748]     chrY 59338754-59338859      + |
#>   [459749]     chrY 59338754-59338859      + |
#>   [459750]     chrY 59340194-59340278      + |
#>   [459751]     chrY 59342487-59343488      + |
#>   [459752]     chrY 59342487-59343488      + |
#>                                          name     score
#>                                   <character> <numeric>
#>        [1]    NR_046018_exon_0_0_chr1_11874_f         0
#>        [2]    NR_046018_exon_1_0_chr1_12613_f         0
#>        [3]    NR_046018_exon_2_0_chr1_13221_f         0
#>        [4]    NR_024540_exon_0_0_chr1_14362_r         0
#>        [5]    NR_024540_exon_1_0_chr1_14970_r         0
#>        ...                                ...       ...
#>   [459748] NM_002186_exon_6_0_chrY_59338754_f         0
#>   [459749] NM_176786_exon_7_0_chrY_59338754_f         0
#>   [459750] NM_002186_exon_7_0_chrY_59340194_f         0
#>   [459751] NM_002186_exon_8_0_chrY_59342487_f         0
#>   [459752] NM_176786_exon_8_0_chrY_59342487_f         0
#>   -------
#>   seqinfo: 93 sequences from an unspecified genome; no seqlengths
exons %>%
  group_by(seqnames) %>%
  summarise(n = n())
#> DataFrame with 49 rows and 2 columns
#>                 seqnames         n
#>                    <Rle> <integer>
#> 1                   chr1     43366
#> 2   chr1_gl000191_random        42
#> 3   chr1_gl000192_random        46
#> 4                  chr10     19420
#> 5                  chr11     24476
#> ...                  ...       ...
#> 45        chrUn_gl000222        20
#> 46        chrUn_gl000223        22
#> 47        chrUn_gl000228        85
#> 48                  chrX     18173
#> 49                  chrY      4128

Next we create a column representing the transcript_id with mutate:

exons <- exons %>%
  mutate(tx_id = sub("_exon.*", "", name))

To find all GWAS SNPs that overlap exons, we use join_overlap_inner. This will create a new GRanges with the coordinates of SNPs that overlap exons, as well as metadata from both objects.

olap <- join_overlap_inner(gwas, exons)
olap
#> GRanges object with 3439 ranges and 4 metadata columns:
#>          seqnames    ranges strand |      name.x
#>             <Rle> <IRanges>  <Rle> | <character>
#>      [1]     chr1   1079198      * |  rs11260603
#>      [2]     chr1   1247494      * |     rs12103
#>      [3]     chr1   1247494      * |     rs12103
#>      [4]     chr1   1247494      * |     rs12103
#>      [5]     chr1   1247494      * |     rs12103
#>      ...      ...       ...    ... .         ...
#>   [3435]     chrX 153764217      * |   rs1050828
#>   [3436]     chrX 153764217      * |   rs1050828
#>   [3437]     chrX 153764217      * |   rs1050828
#>   [3438]     chrX 153764217      * |   rs1050828
#>   [3439]     chrX 153764217      * |   rs1050828
#>                                          name.y     score        tx_id
#>                                     <character> <numeric>  <character>
#>      [1]      NR_038869_exon_2_0_chr1_1078119_f         0    NR_038869
#>      [2]   NM_001256456_exon_1_0_chr1_1247398_r         0 NM_001256456
#>      [3]   NM_001256460_exon_1_0_chr1_1247398_r         0 NM_001256460
#>      [4]   NM_001256462_exon_1_0_chr1_1247398_r         0 NM_001256462
#>      [5]   NM_001256463_exon_1_0_chr1_1247398_r         0 NM_001256463
#>      ...                                    ...       ...          ...
#>   [3435] NM_001042351_exon_9_0_chrX_153764152_r         0 NM_001042351
#>   [3436]    NM_000402_exon_9_0_chrX_153764152_r         0    NM_000402
#>   [3437] NM_001042351_exon_9_0_chrX_153764152_r         0 NM_001042351
#>   [3438]    NM_000402_exon_9_0_chrX_153764152_r         0    NM_000402
#>   [3439] NM_001042351_exon_9_0_chrX_153764152_r         0 NM_001042351
#>   -------
#>   seqinfo: 93 sequences from an unspecified genome; no seqlengths

For each SNP we can count the number of times it overlaps a transcript.

olap %>%
  group_by(name.x, tx_id) %>%
  summarise(n = n())
#> DataFrame with 1619 rows and 3 columns
#>           name.x        tx_id         n
#>      <character>  <character> <integer>
#> 1     rs10043775 NM_001271723         1
#> 2     rs10043775    NM_030793         1
#> 3        rs10078 NM_001242412         1
#> 4        rs10078    NM_020731         1
#> 5        rs10089    NM_001046         1
#> ...          ...          ...       ...
#> 1615   rs9906595 NM_001008777         1
#> 1616      rs9948    NM_017623         1
#> 1617      rs9948    NM_199078         1
#> 1618    rs995030    NM_000899         4
#> 1619    rs995030    NM_003994         4

We can also generate 2bp splice sites on either side of the exon using flank_left and flank_right. We add a column indicating the side of flanking for illustrative purposes. The interweave function pairs the left and right ranges objects.

left_ss <- flank_left(exons, 2L)
right_ss <- flank_right(exons, 2L)
all_ss <- interweave(left_ss, right_ss, .id = "side")
all_ss
#> GRanges object with 919504 ranges and 4 metadata columns:
#>            seqnames            ranges strand |
#>               <Rle>         <IRanges>  <Rle> |
#>        [1]     chr1       11872-11873      + |
#>        [2]     chr1       12228-12229      + |
#>        [3]     chr1       12611-12612      + |
#>        [4]     chr1       12722-12723      + |
#>        [5]     chr1       13219-13220      + |
#>        ...      ...               ...    ... .
#>   [919500]     chrY 59340279-59340280      + |
#>   [919501]     chrY 59342485-59342486      + |
#>   [919502]     chrY 59343489-59343490      + |
#>   [919503]     chrY 59342485-59342486      + |
#>   [919504]     chrY 59343489-59343490      + |
#>                                          name     score       tx_id        side
#>                                   <character> <numeric> <character> <character>
#>        [1]    NR_046018_exon_0_0_chr1_11874_f         0   NR_046018        left
#>        [2]    NR_046018_exon_0_0_chr1_11874_f         0   NR_046018       right
#>        [3]    NR_046018_exon_1_0_chr1_12613_f         0   NR_046018        left
#>        [4]    NR_046018_exon_1_0_chr1_12613_f         0   NR_046018       right
#>        [5]    NR_046018_exon_2_0_chr1_13221_f         0   NR_046018        left
#>        ...                                ...       ...         ...         ...
#>   [919500] NM_002186_exon_7_0_chrY_59340194_f         0   NM_002186       right
#>   [919501] NM_002186_exon_8_0_chrY_59342487_f         0   NM_002186        left
#>   [919502] NM_002186_exon_8_0_chrY_59342487_f         0   NM_002186       right
#>   [919503] NM_176786_exon_8_0_chrY_59342487_f         0   NM_176786        left
#>   [919504] NM_176786_exon_8_0_chrY_59342487_f         0   NM_176786       right
#>   -------
#>   seqinfo: 93 sequences from an unspecified genome; no seqlengths

Learning more

Citation

If you found plyranges useful for your work please cite our paper:

@ARTICLE{Lee2019,
  title    = "plyranges: a grammar of genomic data transformation",
  author   = "Lee, Stuart and Cook, Dianne and Lawrence, Michael",
  journal  = "Genome Biol.",
  volume   =  20,
  number   =  1,
  pages    = "4",
  month    =  jan,
  year     =  2019,
  url      = "http://dx.doi.org/10.1186/s13059-018-1597-8",
  doi      = "10.1186/s13059-018-1597-8",
  pmc      = "PMC6320618"
}

Contributing

We welcome contributions from the R/Bioconductor community. We ask that contributors follow the code of conduct and the guide outlined here.