BiocStyle::markdown()

Package: Pbase
Author: Johannes Rainer, Laurent Gatto
Last compiled: r date()
Last modified: r file.info("Pbase-with-ensembldb.Rmd")$mtime

library("Pbase")
library("ensembldb")

Introduction

This vignette describes how the r Biocpkg("ensembldb") package can be used to retrieve protein annotations and to map peptides within protein sequences to the genomic coordinates. EnsDb databases and packages created by the ensembldb package version > 1.99.0 can provide protein annotations, but don't necessarily have too. Only EnsDb packages created using the Ensembl Perl API contain protein annotations, while databases created from GTF or GFF files or from GRanges objects or using the r Biocpkg("AnnotationHub") don't.

Below we load the EnsDb object with for all human genes defined in the Ensembl database version 86 and use the hasProteinData method to evaluate whether protein annotations are available.

library(ensembldb)
library(EnsDb.Hsapiens.v86)
## Make a shortcut to the object
edb <- EnsDb.Hsapiens.v86
hasProteinData(edb)

Fetch protein annotations from the database

ensembldb provides the proteins method to fetch protein annotations from an EnsDb database (along eventual transcript or gene annotations) and, depending on the value of the return.type parameter, return the results as a data.frame, DataFrame or AAStringSet. Below we use the function to retrieve all proteins encoded by the gene ZBTB16. We use a GeneNameFilter to select entries for that gene. In the simple example below only protein annotations were retrieved (i.e. the Ensembl protein ID, the amino acid sequence, the ID of the transcript encoding the protein and the gene name; the latter two stored in the AAStringSet's mcols), but EnsDbs provide in addition also the mapping between Ensembl protein IDs and Uniprot IDs and all protein domains within the protein sequence. For more information on available filters, methods and annotation columns see the ensembldb vignette.

prts <- proteins(edb, filter = GeneNameFilter("ZBTB16"),
                 return.type = "AAStringSet")
prts

## Get access to the additional annotation data
mcols(prts)

In Pbase, a Proteins method is also implemented for EnsDb objects that enables to load a Proteins object from an EnsDb database and optionally load all protein domains as peptide ranges in the pranges slot (using the loadProteinDomains parameter which is by default TRUE). Below we use also a filter expression in form of a formula instead of an AnnotationFilter class.

library(Pbase)
zbtb16 <- Proteins(edb, filter = ~ gene_name == "ZBTB16")

The proteins' sequences are stored in the aa slot and the protein domains in the pranges slot:

## Get the protein sequences.
aa(zbtb16)

## Get the protein domains stored as peptide features.
pranges(zbtb16)

Also the Proteins method supports to retrieve additional annotation columns from the database. Below we repeat the call but fetch in addition also the Uniprot IDs for the proteins. With that a property of the provided annotation becomes apparent: while the mapping between Ensembl transcript ID and Ensembl protein ID is always 1:1, each Ensembl protein can be annotated to more than one Uniprot ID and each Uniprot ID can be assigned one or more Ensembl protein IDs.

## Loading in addition Uniprot ID annotations
zbtb16_2 <- Proteins(edb, columns = c("uniprot_id", "uniprot_db",
                                      "uniprot_mapping_type"),
                     filter = GeneNameFilter("ZBTB16"))

acols(zbtb16_2)

As we can see two transcript IDs (ENST00000335953 and ENST00000392996) and hence Ensembl protein IDs are each annotated to two Uniprot IDs. To avoid some of these multi-mappings, we can use an UniprotMappingTypeFilter to retrieve only presumably higher quality annotations by selecting only Uniprot IDs mapped to Ensembl protein IDs using the "DIRECT" mapping type. We get hence again unique Ensembl protein IDs respectively transcript IDs.

acols(Proteins(edb, columns = c("uniprot_id", "uniprot_db"),
               filter = AnnotationFilterList(
                   GeneNameFilter("ZBTB16"),
                   UniprotMappingTypeFilter("DIRECT"))))

Note that without specifying a filter we can also retrieve all proteins from the database.

Protein and genome data

EnsDb databases provide both, protein annotations and gene/transcript annotations including their genomic coordinates. Peptide features along amino acid sequences within a Proteins objects can hence be mapped to the genome directly using the mapToGenome method by providing an EnsDb instance with the genome parameter. The Proteins object has to provide IDs that allow to identify the encoding transcripts. Supported IDs are Ensembl protein ID (idType = "protein_id"), transcript ID (idType = "tx_id") or Uniprot ID (idType = "uniprot_id") which can be provided either as the names of the Proteins object, or in one of the acols metadata columns.

In the example below we use the Proteins object with all proteins of the gene ZBTB16 that we fetched from the database in the previous section. To identify the encoding transcripts, we use the Ensembl protein IDs that are provided with the names of the object.

## We use the Ensembl protein IDs to identify the transcripts in the database
names(zbtb16)

## Map all peptide features within the object to the genome.
zbtb16_map <- mapToGenome(zbtb16, edb, idType = "protein_id", id = "name")

zbtb16_map

The result is a GRangesList one element for each protein with GRanges for the genomic positions of the mapped peptide features (in our example protein domains within the protein sequence). Below we plot the genomic alignments for the protein domains of the first protein.

library(Gviz)

## Define a genome axis track
gat <- GenomeAxisTrack()

## Get the transcript ID of the first transcript:
txid <- acols(zbtb16)$tx_id[1]
## Get a GRanges for the first transcript
trt <- getGeneRegionTrackForGviz(edb, filter = TxIdFilter(txid))

## Add the protein domain name as column "id" to the mcols so it
## can be used in the plot to identify the features.
map_1 <- zbtb16_map[[1]]
map_1$id <- names(map_1)

## Plotting the transcript and the mapped protein domains.
plotTracks(list(gat,
                GeneRegionTrack(trt, name = "tx"),
                AnnotationTrack(map_1, groupAnnotation = "id",
                                just.group = "above",
                                name = "Protein domains")),
           transcriptAnnotation = "transcript")

While the mapToGenome,Proteins,EnsDb method used above performs all of the mapping and supports mapping of multiple proteins, we could also fetch the genomic positions of the encoding transcripts' CDS using the cdsBy method and provide the resulting GRangesList to the mapToGenome,Proteins,GRangesList method.

In addition to Ensembl transcript or protein IDs, it is also possible to provide Uniprot IDs. While the mapping between Ensembl protein IDs and transcript IDs is 1:1, multiple Ensembl protein IDs can be assigned to a single Uniprot ID hence the mappoing between Uniprot IDs and Ensembl transcript IDs is also 1:n. In cases in which a Uniprot ID is annotated to multiple transcript IDs, the mapToGenome method will select automatically the best suited transcript by comparing the length of the protein sequence with the length of the transcripts' coding sequences. In the example below we use the test Proteins data from the Pbase package and perform the mapping of the peptide features using the provided Uniprot IDs of the proteins.

data(p)

## We use the Uniprot IDs provided as names for the mapping
names(p)

res <- mapToGenome(p, edb, idType = "uniprot_id")

Apart from one of the Uniprot IDs, P04075-2, for which no corresponding transcript could be identified, all peptide features could be mapped successfully. As detailed above, the method did select for each Uniprot ID the transcript with the best matching CDS. To illustrate this, we select below all transcripts that are annotated to the first Uniprot ID.

txs <- transcripts(edb, filter = UniprotFilter(names(p)[1]))
names(txs)

While there are 6 transcripts annotated to the Uniprot ID from only one the coding sequence matches the length of the protein:

## Fetch the CDS for each transcript
cdss <- cdsBy(edb, filter = TxIdFilter(names(txs)))

## The protein sequence length:
width(aa(p)[1])

## The difference between protein sequence and CDS length
width(aa(p)[1]) - sum(width(cdss)) / 3

Thus, the lenght of the coding region of the last transcript matches the length of the protein's amino acid sequence (the stop codon, being part of the CDS, is not coding for an amino acid).

Session information

sessionInfo()


ComputationalProteomicsUnit/Pbase documentation built on Aug. 10, 2019, 1:25 a.m.