This documents describes two use cases for the coordinate system mapping functionality of ensembldb: mapping of regions within protein sequences to the genome and mapping of genomic to protein sequence-relative coordinates. In addition, it showcases the advanced filtering capabilities implemented in ensembldb.

1 Query for helix-loop-helix transcription factors on chromosome 21

Down syndrome is a genetic disorder characterized by the presence of all or parts of a third copy of chromosome 21. It is associated, among other, with characteristic facial features and mild to moderate intellectual disability. The phenotypes are most likely the result from a gene dosage-dependent increased expression of the genes encoded on chromosome 21 (Lana-Elola et al. 2011). Compared to other gene classes, transcription factors are more likely to have an immediate impact, even due to a moderate over-expression (which might be the result from gene duplication). One of the largest dimerizing transcription factor families is characterized by a basic helix-loop-helix domain (Massari and Murre 2000), a protein structural motif facilitating DNA binding.

The example below aims at identifying transcription factors with a basic helix-loop-helix domain (Pfam ID PF00010) that are encoded on chromosome 21. To this end we first load an R-library providing human annotations from Ensembl release 86 and pass the loaded EnsDb object along with a filter expression to the genes method that retrieves the corresponding genes. Filter expressions have to be written in the form ~ <field> <condition> <value> with <field> representing the database column to be used for the filter. Several such filter expressions can be concatenated with standard R logical expressions (such as & or |). To get a list of all available filters and their corresponding fields, the supportedFilters(edb) function could be used.

library(ensembldb)
library(EnsDb.Hsapiens.v86)
edb <- EnsDb.Hsapiens.v86

## Retrieve the genes
gns <- genes(edb, filter = ~ protein_domain_id == "PF00010" & seq_name == "21")

The function returned a GRanges object with the genomic position of the genes and additional gene-related annotations stored in metadata columns.

gns
## GRanges object with 3 ranges and 7 metadata columns:
##                   seqnames            ranges strand |         gene_id
##                      <Rle>         <IRanges>  <Rle> |     <character>
##   ENSG00000205927       21 33025845-33029196      + | ENSG00000205927
##   ENSG00000184221       21 33070144-33072420      + | ENSG00000184221
##   ENSG00000159263       21 36699133-36749917      + | ENSG00000159263
##                     gene_name   gene_biotype seq_coord_system      symbol
##                   <character>    <character>      <character> <character>
##   ENSG00000205927       OLIG2 protein_coding       chromosome       OLIG2
##   ENSG00000184221       OLIG1 protein_coding       chromosome       OLIG1
##   ENSG00000159263        SIM2 protein_coding       chromosome        SIM2
##                   entrezid protein_domain_id
##                     <list>       <character>
##   ENSG00000205927    10215           PF00010
##   ENSG00000184221   116448           PF00010
##   ENSG00000159263     6493           PF00010
##   -------
##   seqinfo: 1 sequence from GRCh38 genome

Three transcription factors with a helix-loop-helix domain are encoded on chromosome 21: SIM2, which is a master regulator of neurogenesis and is thought to contribute to some specific phenotypes of Down syndrome (Gardiner and Costa 2006) and the two genes OLIG1 and OLIG2 for which genetic triplication was shown to cause developmental brain defects (Chakrabarti et al. 2010). To visualize the exonic regions encoding the helix-loop-helix domain of these genes we next retrieve their transcript models and the positions of all Pfam protein domains within the amino acid sequences of encoded by these transcripts. We process SIM2 separately from OLIG1 and OLIG2 because the latter are encoded in a narrow region on chromosome 21 and can thus be visualized easily within the same plot. We extract the transcript models for OLIG1 and OLIG2 that encode the protein domain using the getGeneRegionTrackForGviz function which returns the data in a format that can be directly passed to functions from the Gviz Bioconductor package (Hahne and Ivanek 2016) for plotting. Since Gviz expects UCSC-style chromosome names instead of the Ensembl chromosome names (e.g. chr21 instead of 21), we change the format in which chromosome names are returned by ensembldb with the seqlevelsStyle method. All subsequent queries to the EnsDb database will return chromosome names in UCSC format.

## Change chromosome naming style to UCSC
seqlevelsStyle(edb) <- "UCSC"
## Retrieve the transcript models for OLIG1 and OLIG2 that encode the
## the protein domain
txs <- getGeneRegionTrackForGviz(
    edb, filter = ~ genename %in% c("OLIG1", "OLIG2") &
             protein_domain_id == "PF00010")

Next we fetch the coordinates of all Pfam protein domains encoded by these transcripts with the proteins method, asking for columns "prot_dom_start", "prot_dom_end" and "protein_domain_id" to be returned by the function. Note that we restrict the results in addition to protein domains defined in Pfam.

pdoms <- proteins(edb, filter = ~ tx_id %in% txs$transcript &
                           protein_domain_source == "pfam",
                  columns = c("protein_domain_id", "prot_dom_start",
                              "prot_dom_end"))
pdoms
## DataFrame with 3 rows and 6 columns
##   protein_domain_id prot_dom_start prot_dom_end      protein_id           tx_id
##         <character>      <integer>    <integer>     <character>     <character>
## 1           PF00010            107          164 ENSP00000371785 ENST00000382348
## 2           PF00010            110          162 ENSP00000371794 ENST00000382357
## 3           PF00010            110          162 ENSP00000331040 ENST00000333337
##   protein_domain_source
##             <character>
## 1                  pfam
## 2                  pfam
## 3                  pfam

We next map these protein-relative positions to the genome. We define first an IRanges object with the coordinates and submit this to the proteinToGenome function for mapping. Besides coordinates, the function requires also the respective protein identifiers which we supply as names.

pdoms_rng <- IRanges(start = pdoms$prot_dom_start, end = pdoms$prot_dom_end,
                     names = pdoms$protein_id)

pdoms_gnm <- proteinToGenome(pdoms_rng, edb)

The result is a list of GRanges objects with the genomic coordinates at which the protein domains are encoded, one for each of the input protein domains. Additional information such as the protein ID, the encoding transcript and the exons of the respective transcript in which the domain is encoded are provided as metadata columns.

pdoms_gnm
## $ENSP00000371785
## GRanges object with 1 range and 7 metadata columns:
##       seqnames            ranges strand |      protein_id           tx_id
##          <Rle>         <IRanges>  <Rle> |     <character>     <character>
##   [1]    chr21 33070565-33070738      + | ENSP00000371785 ENST00000382348
##               exon_id exon_rank    cds_ok protein_start protein_end
##           <character> <integer> <logical>     <integer>   <integer>
##   [1] ENSE00001491811         1      TRUE           107         164
##   -------
##   seqinfo: 1 sequence from GRCh38 genome
## 
## $ENSP00000371794
## GRanges object with 1 range and 7 metadata columns:
##       seqnames            ranges strand |      protein_id           tx_id
##          <Rle>         <IRanges>  <Rle> |     <character>     <character>
##   [1]    chr21 33027190-33027348      + | ENSP00000371794 ENST00000382357
##               exon_id exon_rank    cds_ok protein_start protein_end
##           <character> <integer> <logical>     <integer>   <integer>
##   [1] ENSE00001491833         2      TRUE           110         162
##   -------
##   seqinfo: 1 sequence from GRCh38 genome
## 
## $ENSP00000331040
## GRanges object with 1 range and 7 metadata columns:
##       seqnames            ranges strand |      protein_id           tx_id
##          <Rle>         <IRanges>  <Rle> |     <character>     <character>
##   [1]    chr21 33027190-33027348      + | ENSP00000331040 ENST00000333337
##               exon_id exon_rank    cds_ok protein_start protein_end
##           <character> <integer> <logical>     <integer>   <integer>
##   [1] ENSE00001379164         1      TRUE           110         162
##   -------
##   seqinfo: 1 sequence from GRCh38 genome

Column cds_ok in the result object indicates whether the length of the CDS of the encoding transcript matches the length of the protein sequence. For transcripts with unknown 3’ and/or 5’ CDS ends these will differ. The mapping result has to be re-organized before being plotted: Gviz expects a single GRanges object, with specific metadata columns for the grouping of the individual genomic regions. This is performed in the code block below.

## Convert the list to a GRanges with grouping information
pdoms_gnm_grng <- unlist(GRangesList(pdoms_gnm))
pdoms_gnm_grng$id <- rep(pdoms$protein_domain_id, lengths(pdoms_gnm))
pdoms_gnm_grng$grp <- rep(1:nrow(pdoms), lengths(pdoms_gnm))

pdoms_gnm_grng
## GRanges object with 3 ranges and 9 metadata columns:
##                   seqnames            ranges strand |      protein_id
##                      <Rle>         <IRanges>  <Rle> |     <character>
##   ENSP00000371785    chr21 33070565-33070738      + | ENSP00000371785
##   ENSP00000371794    chr21 33027190-33027348      + | ENSP00000371794
##   ENSP00000331040    chr21 33027190-33027348      + | ENSP00000331040
##                             tx_id         exon_id exon_rank    cds_ok
##                       <character>     <character> <integer> <logical>
##   ENSP00000371785 ENST00000382348 ENSE00001491811         1      TRUE
##   ENSP00000371794 ENST00000382357 ENSE00001491833         2      TRUE
##   ENSP00000331040 ENST00000333337 ENSE00001379164         1      TRUE
##                   protein_start protein_end          id       grp
##                       <integer>   <integer> <character> <integer>
##   ENSP00000371785           107         164     PF00010         1
##   ENSP00000371794           110         162     PF00010         2
##   ENSP00000331040           110         162     PF00010         3
##   -------
##   seqinfo: 1 sequence from GRCh38 genome

We next define the individual tracks we want to visualize and plot them with the plotTracks function from the Gviz package.

library(Gviz)

## Define the individual tracks:
## - Ideogram
## ideo_track <- IdeogramTrack(genome = "hg38", chromosome = "chr21")
## - Genome axis
gaxis_track <- GenomeAxisTrack()
## - Transcripts
gene_track <- GeneRegionTrack(txs, showId = TRUE, just.group = "right",
                              name = "", geneSymbol = TRUE, size = 0.5)
## - Protein domains
pdom_track <- AnnotationTrack(pdoms_gnm_grng, group = pdoms_gnm_grng$grp,
                              id = pdoms_gnm_grng$id, groupAnnotation = "id",
                              just.group = "right", shape = "box",
                              name = "Protein domains", size = 0.5)

## Generate the plot
plotTracks(list(gaxis_track, gene_track, pdom_track))
Transcripts of genes OLIG1 and OLIG2 encoding a helix-loop-helix protein domain. Shown are all transcripts of the genes OLIG1 and OLIG2 that encode a protein with a helix-loop-helix protein domain (PF00010). Genomic positions encoding protein domains defined in Pfam are shown in light blue.

Figure 1: Transcripts of genes OLIG1 and OLIG2 encoding a helix-loop-helix protein domain
Shown are all transcripts of the genes OLIG1 and OLIG2 that encode a protein with a helix-loop-helix protein domain (PF00010). Genomic positions encoding protein domains defined in Pfam are shown in light blue.

All transcripts are relatively short with the full coding region being in a single exon. Also, both transcripts encode a protein with a single protein domain, the helix-loop-helix domain PF00010.

Next we repeat the analysis for SIM2 by first fetching all of its transcript variants encoding the PF00010 Pfam protein domain from the database. Subsequently we retrieve all Pfam protein domains encoded in these transcripts.

## Fetch all SIM2 transcripts encoding PF00010
txs <- getGeneRegionTrackForGviz(edb, filter = ~ genename == "SIM2" &
                                          protein_domain_id == "PF00010")
## Fetch all Pfam protein domains within these transcripts
pdoms <- proteins(edb, filter = ~ tx_id %in% txs$transcript &
                           protein_domain_source == "pfam",
                  columns = c("protein_domain_id", "prot_dom_start",
                              "prot_dom_end"))

At last we have to map the protein domain coordinates to the genome and prepare the data for the plot. Since the code is essentially identical to the one for OLIG1 and OLIG2 it is not displayed.

Transcripts of the gene SIM2 encoding a helix-loop-helix domain. Shown are all transcripts of SIM2 encoding a protein with a helix-loop-helix protein domain (PF00010). Genomic positions encoding protein domains defined in Pfam are shown in light blue.

Figure 2: Transcripts of the gene SIM2 encoding a helix-loop-helix domain
Shown are all transcripts of SIM2 encoding a protein with a helix-loop-helix protein domain (PF00010). Genomic positions encoding protein domains defined in Pfam are shown in light blue.

The SIM2 transcript encodes a protein with in total 4 protein domains. The helix-loop-helix domain PF00010 is encoded in its first exon.

2 Mapping of genomic coordinates to protein-relative positions

One of the known mutations for human red hair color is located at position 16:89920138 (dbSNP ID rs1805009) on the human genome (version GRCh38). Below we map this genomic coordinate to the respective coordinate within the protein sequence encoded at that location using the genomeToProtein function. Note that we use "chr16" as the name of the chromosome, since we changed the chromosome naming style to UCSC in the previous example.

gnm_pos <- GRanges("chr16", IRanges(89920138, width = 1))
prt_pos <- genomeToProtein(gnm_pos, edb)
prt_pos
## IRangesList object of length 1:
## [[1]]
## IRanges object with 3 ranges and 8 metadata columns:
##                       start       end     width |           tx_id    cds_ok
##                   <integer> <integer> <integer> |     <character> <logical>
##   ENSP00000451605       294       294         1 | ENST00000555147      TRUE
##   ENSP00000451760       294       294         1 | ENST00000555427      TRUE
##   ENSP00000451560       294       294         1 | ENST00000556922      TRUE
##                           exon_id exon_rank seq_start   seq_end    seq_name
##                       <character> <integer> <integer> <integer> <character>
##   ENSP00000451605 ENSE00002458332         1  89920138  89920138       chr16
##   ENSP00000451760 ENSE00002477640         3  89920138  89920138       chr16
##   ENSP00000451560 ENSE00002507842         1  89920138  89920138       chr16
##                    seq_strand
##                   <character>
##   ENSP00000451605           *
##   ENSP00000451760           *
##   ENSP00000451560           *

The genomic position could thus be mapped to the amino acid 294 in each of the 3 proteins listed above. Using the select function we retrieve the official symbol of the gene for these 3 proteins.

select(edb, keys = ~ protein_id == names(prt_pos[[1]]), columns = "SYMBOL")
##          SYMBOL       PROTEINID
## 1 RP11-566K11.2 ENSP00000451560
## 2          MC1R ENSP00000451605
## 3          MC1R ENSP00000451760

Two proteins are from the MC1R gene and one from RP11-566K11.2 (ENSG00000198211) a gene which exons overlap exons from MC1R as well as exons of the more downstream located gene TUBB3. To visualize this we first fetch transcripts overlapping the genomic position of interest and subsequently all additional transcripts within the region defined by the most downstream and upstream exons of the transcripts.

## Get transcripts overlapping the genomic position.
txs <- getGeneRegionTrackForGviz(edb, filter = GRangesFilter(gnm_pos))

## Get all transcripts within the region from the start of the most 5'
## and end of the most 3' exon.
all_txs <- getGeneRegionTrackForGviz(
    edb, filter = GRangesFilter(range(txs), type = "within"))

## Plot the data
## - Ideogram
## ideo_track <- IdeogramTrack(genome = "hg38", chromosome = "chr16")
## - Genome axis
gaxis_track <- GenomeAxisTrack()
## - Transcripts
gene_track <- GeneRegionTrack(all_txs, showId = TRUE, just.group = "right",
                              name = "", geneSymbol = TRUE, size = 0.5)
## - highlight the region.
hl_track <- HighlightTrack(list(gaxis_track, gene_track), range = gnm_pos)

## Generate the plot
plotTracks(list(hl_track))