This package and the underlying code are distributed under the Artistic license 2.0. You are free to use and redistribute this software.

1 Rationale

“The ENCODE (Encyclopedia of DNA Elements) Consortium is an international collaboration of research groups funded by the National Human Genome Research Institute (NHGRI). The goal of ENCODE is to build a comprehensive parts list of functional elements in the human genome, including elements that act at the protein and RNA levels, and regulatory elements that control cells and circumstances in which a gene is active” source: ENCODE Projet Portal.

However, retrieving and downloading data can be time consuming using the current web portal, especially when multiple files from different experiments are involved.

This package has been designed to facilitate access to ENCODE data by compiling the metadata associated with files, experiments, datasets, biosamples, and treatments.

We implemented time-saving features to select ENCODE files by querying their metadata, downloading them and validating that the file was correctly downloaded.

This vignette will introduce the main features of the ENCODExplorer package.

2 Loading ENCODExplorer package

library(ENCODExplorer)

3 Introduction

To use the functionalities of the ENCODExplorer package, you first need to download the data.table containing all the metadata from ENCODE.

This data.table is available through the AnnotationHub package. For convenience, the latest available version at each release will be downloaded and used by default.

We also provide the following function to quickly extract those metadata:

To load getencode_df :

encode_df <- get_encode_df()
## snapshotDate(): 2019-05-01
## downloading 0 resources
## loading from cache 
##     'AH69290 : 76036'

4 Main functions

4.1 Query

The queryEncode function allows the user to find the subset of files corresponding to a precise query defined according to the following criteria :

Parameter	Description
set_accession	The accession for the containing experiment or dataset
dataset_accession	There is a subtle difference between the parameters set_accession and dataset_accession. In fact, some files can be part of an experiment, a dataset or both. When using set_accession, you will get all the files directly associated with this accession (experiment and/or dataset). While the usage of dataset_accession will get the files directly associated to the requested dataset AND those which are part of an experiment and indirectly linked to a dataset (reported as related files in the dataset and related_dataset in the experiment).
file_accession	The accesion for one specific file
biosample_name	The biosample name (“GM12878”, “kidney”)
biosample_type	The biosample type (“tissue”, “cell line”)
assay	The assay type (“ChIP-seq”, “polyA RNA-seq”)
file_format	The file format. Some currently available formats include bam, bed, fastq, bigBed, bigWig, CEL, csfasta, csqual, fasta, gff, gtf, idat, rcc, sam, tagAlign, tar, tsv, vcf, wig.
lab	The laboratory
organism	The donor organism (“Homo sapiens”, “Mus musculus”)
target	The gene, protein or histone mark which was targeted by the assay (Immunoprecipitated protein in ChIP-seq, knocked-down gene in CRISPR RNA-seq assays, etc)
treatment	The treatment related to the biosample
project	The project name/id

By default, the query function uses exact string matching to perform the selection of the relevant entries. This behavior can be changed by changing the fixed or fuzzy parameters. Setting fixed to FALSE will perform case-insensitive regular expression matching. Setting fuzzy to TRUE will retrieve search results where the query string is only a partial match.

The result set is a subset of the encode_df_lite table.

For example, to select all the fastq files originating from assays on the MCF-7 (human breast cancer) cell line:

query_results <- queryEncode(organism = "Homo sapiens", 
                      biosample_name = "MCF-7", file_format = "fastq",
                      fixed = TRUE)
## Results : 813 files, 234 datasets

The same request with approximate spelling of the biosample name and fuzzy option to FALSE, will give no results :

query_results <- queryEncode(organism = "Homo sapiens", biosample_name = "mcf7",
                        file_format = "fastq", fixed = TRUE,
                       fuzzy = FALSE)
## No result found in encode_df. You can try the <searchEncode> function or set the fuzzy option to TRUE.

If you follow the warning guidance and set the fuzzy option to TRUE:

query_results <- queryEncode(organism = "Homo sapiens",
                    biosample_name = "mcf7", file_format = "fastq", fixed = TRUE,
                    fuzzy = TRUE)
## Results : 813 files, 234 datasets

You can also perform matching through regular expressions by setting fixed to FALSE.

query_results <- queryEncode(assay = ".*RNA-seq",
                    biosample_name = "HeLa-S3", fixed = FALSE)
## Results : 318 files, 11 datasets
table(query_results$assay)
## 
##          polyA RNA-seq polyA depleted RNA-seq          small RNA-seq 
##                    150                     90                     78

Finally, the queryEncodeGeneric function can be used to perform searches on columns which are not part of the queryEncode interface but are present within the encode_df_lite data.table:

query_results <- queryEncodeGeneric(biosample_name="HeLa-S3",
                    assay="RNA-seq", submitted_by="Diane Trout",
                    fuzzy=TRUE)
## Results : 54 files, 2 datasets
table(query_results$submitted_by)
## 
## Diane Trout 
##          54

These criteria correspond to the filters that you can find on ENCODE portal :

results of a filtered search on ENCODE portal

4.2 fuzzySearch

This function is a more user-friendly version of queryEncode that also searches on the encode_df_lite object. The character vector or the list of characters specified by the user will be searched for in every column of the database. The user can also constrain the query by selecting the specific column in which to search for the query term by using the filterVector parameter.

The following request will produce a data.table with every files containing the term brca.

fuzzy_results <- fuzzySearch(searchTerm = c("brca"))
## Results: 236 files, 7 datasets

Multiple terms can be searched simultaneously. This example extracts all files containing brca or ZNF24 within the target column.

fuzzy_results <- fuzzySearch(searchTerm = c("brca", "ZNF24"),
                             filterVector = c("target"),
                             multipleTerm = TRUE)
## Results: 710 files, 17 datasets

When searching for multiple terms, three type of input can be passed to the searchTerm parameter : - A single character where the various terms are separated by commas - A character vector - A list of characters

4.3 Search

This function simulates a keyword search performed through the ENCODE web portal.

The searchEncode function returns a data frame corresponding to the result page provided by the ENCODE portal. If a specific file or dataset isn’t available with fuzzySearch or queryEncode (i.e. within get_encode_df()), the user can access the latest data from the ENCODE database through the searchEncode function.

The searchToquery function convert the result of a search to a data.table with the same design as get_encode_df(). This format contains more metadata and allow the user to extract all files within the dataset. This format also allows the user to create a design using the createDesign function.

Here is the example of the following search : “a549 chip-seq homo sapiens”.

On ENCODE portal :

results of a key word search on ENCODE portal

With our function :

  search_results <- searchEncode(searchTerm = "a549 chip-seq homo sapiens",
                                 limit = "all")
## results : 414

4.4 createDesign

This function organizes the data.table created by fuzzySearch, queryEncode or searchToquery. It extracts the replicate and control files within a dataset.

It creates a data.table with the file accessions, the dataset accessions and numeric values associated with the nature of the file (1:replicate / 2:control) when the format parameter is set to long.

By setting the format parameter to wide, each dataset will have its own column as illustrated below.

Wide design example

4.5 downloadEncode

downloadEncode allows a user to download a file or an entire dataset. Downloading files can be done by providing a vector of file accessions or dataset accessions (represented by the accession column in get_encode_df()) to the file_acc parameter. This parameter can also be the data.table created by queryEncode, fuzzySearch, searchToquery or createDesign.

If the accession doesn’t exist within the passed-in get_encode_df() database, downloadEncode will search for the accession directly within the ENCODE database. The path to the download directory can be specified (default: /tmp).

To ensure the integrity of each file, the md5 sum of each downloaded file is compared to the reported md5 sum in ENCODE.

Moreover, if the accession is a dataset accession, the function will download each file in this dataset. The format option, which is set by default to all, enables the downloading of a specific format.

Here is a small example query:

query_results <- queryEncode(assay = "switchgear", target ="elavl1", fixed = FALSE)
## Results : 2 files, 1 datasets

And its equivalent search:

search_results <- searchEncode(searchTerm = "switchgear elavl1", limit = "all")
## results : 1

To select a particular file format you can:

add filters to your query and then run the downloadEncode function.

query_results <- queryEncode(assay = "switchgear", target ="elavl1",
                             file_format = "bed" , fixed = FALSE)
downloadEncode(query_results)

specify the format to the downloadEncode function.

downloadEncode(search_results, format = "bed")

4.6 Conversion

The function searchToquery enables the conversion of the results of searchEncode to a queryEncode output based on the accession numbers. The user can then benefit from all the collected metadata and the createDesign function.

The structure of the result set is similar to the get_encode_df() structure.

Let’s try it with the previous example :

search_results <- searchEncode(searchTerm = "switchgear elavl1", limit = "all")
## results : 1

convert

convert_results <- searchToquery(searchResults = search_results)

4.7 shinyEncode

This function launches the shinyApp of ENCODExplorer that implements the fuzzySearch and queryEncode search functions. It also allows the creation of a design to organize and download specific files with the downloadEncode function. The Search tab of shinyEncode uses the fuzzySearch function for a low specificity request while the Advanced Search tab uses the queryEncode function.

Simple request using Search

5 Updating the ENCODE file database

By default, ENCODExplorer retrieves the ENCODE metadata from its sister package, ENCODExplorerData. The version of the metadata provided by default will be updated with each Bioconductor release in the ENCODExplorer package. However, since all of ENCODExplorer’s function take an explicit df parameter, it is possible to use the AnnotationHub package to download a more recent version:

require(AnnotationHub)
ah = AnnotationHub()
query(ah, "ENCODExplorerData")

## AnnotationHub with 2 records
## # snapshotDate(): 2019-05-01 
## # $dataprovider: ENCODE Project
## # $species: NA
## # $rdataclass: data.table
## # additional mcols(): taxonomyid, genome, description,
## #   coordinate_1_based, maintainer, rdatadateadded, preparerclass,
## #   tags, rdatapath, sourceurl, sourcetype 
## # retrieve records with, e.g., 'object[["AH69290"]]' 
## 
##             title                                         
##   AH69290 | ENCODE File Metadata (Light, 2019-04-12 build)
##   AH69291 | ENCODE File Metadata (Full, 2019-04-12 build)

Finally, it is also possible to use ENCODExplorerData functionalities to generate an up-to-date data.table, and pass it to ENCODExplorer’s functions.

We refer the user to the ENCODExplorerData vignettes for details on how to generate an up-to-date data.table.

ENCODExplorer: A compilation of metadata from ENCODE

2 May 2019

Contents