Introduction to gtexr

The GTEx Portal API V2 enables programmatic access to data available from the Genotype-Tissue Expression Portal. The gtexr package wraps this API, providing R functions that correspond to each API endpoint:

R function names mirror those of their corresponding endpoint, converted to lower case with spaces replaced with underscores¹ e.g. the R function for “Get Service Info” is get_service_info().
Query parameters are similarly mirrored by function arguments e.g. the arguments for get_maintenance_message() (corresponding to the endpoint “Get Maintenance Message”) are page and itemsPerPage. For query parameters that accept an array of values however, the corresponding function argument is pluralised to indicate this e.g. for endpoint “Get Eqtl Genes” the query parameter ‘tissueSiteDetailId’ is pluralised to argument name tissueSiteDetailIds in get_eqtl_genes().
Default values for arguments mirror those for the API.
The documentation for each function includes at least one working example e.g. ?get_eqtl_genes provides example valid values for the required argument tissueSiteDetailIds.
All functions return a tibble::tibble by default. Alternatively, the raw JSON from an API call may be retrieved by setting argument .return_raw to TRUE e.g. get_service_info(.return_raw = TRUE).

Shiny app

Users can try out all functions interatively with the ⭐gtexr shiny app⭐, which pre-populates query parameters with those for the first working example from each function’s documentation. To run the app locally:

shiny::runApp(system.file("app", package = "gtexr"))

Paginated responses

library(gtexr)
library(dplyr)
library(purrr)

Many API endpoints return only the first 250 available items by default. A warning is raised if the number of available items exceeds the selected maximum page size e.g.

get_eqtl_genes("Whole_Blood")

For most cases, the simplest solution is to increase the value of itemsPerPage e.g. get_eqtl_genes("Whole_Blood", itemsPerPage = 100000). This limit can be set globally by setting the “gtexr.itemsPerPage” option with options(list(gtexr.itemsPerPage = 100000)).

Alternatively, multiple pages can be retrieved sequentially e.g.

# to retrieve the first 3 pages, with default setting of 250 items per page
1:3 |>
  map(\(page) get_eqtl_genes("Whole_Blood", page = page, .verbose = FALSE) |>
        suppressWarnings()) |>
  bind_rows()

Note that paging information is printed to the R console by default. Set argument .verbose to FALSE to silence these messages, or disable globally with options(list(gtexr.verbose = FALSE)).

Examples

The rest of this vignette outlines some example applications of gtexr.

Get build 37 coordinates for a variant

get_variant(snpId = "rs1410858") |>
  tidyr::separate(
    col = b37VariantId,
    into = c(
      "chromosome",
      "position",
      "reference_allele",
      "alternative_allele",
      "genome_build"
    ),
    sep = "_",
    remove = FALSE
  ) |>
  select(snpId:genome_build)

Convert gene symbol to versioned GENCODE ID

Use get_gene() or get_genes()

get_genes("CRP") |>
  select(geneSymbol, gencodeId)

Convert rsID to GTEx variant ID

Use get_variant()

get_variant(snpId = "rs1410858") |>
  select(snpId, variantId)

For a gene of interest, which tissues have significant cis-eQTLs?

Use get_significant_single_tissue_eqtls() (note this requires versioned GENCODE IDs)

gene_symbol_of_interest <- "CRP"

gene_gencodeId_of_interest <- get_genes(gene_symbol_of_interest) |>
  pull(gencodeId) |>
  suppressMessages()

gene_gencodeId_of_interest |>
  get_significant_single_tissue_eqtls() |>
  distinct(geneSymbol, gencodeId, tissueSiteDetailId)

Get data for non-eQTL variants

Some analyses (e.g. Mendelian randomisation) require data for variants which may or may not be significant eQTLs. Use calculate_expression_quantitative_trait_loci() with purrr::map() to retrieve data for multiple variants

variants_of_interest <- c("rs12119111", "rs6605071", "rs1053870")

variants_of_interest |>
  set_names() |>
  map(
    \(x) calculate_expression_quantitative_trait_loci(
      tissueSiteDetailId = "Liver",
      gencodeId = "ENSG00000237973.1",
      variantId = x
    )
  ) |>
  bind_rows(.id = "rsid") |>
  # optionally, reformat output - first extract genomic coordinates and alleles
  tidyr::separate(
    col = "variantId",
    into = c(
      "chromosome",
      "position",
      "reference_allele",
      "alternative_allele",
      "genome_build"
    ),
    sep = "_"
  ) |>
  # ...then ascertain alternative_allele frequency
  mutate(
    alt_allele_count = (2 * homoAltCount) + hetCount,
    total_allele_count = 2 * (homoAltCount + hetCount + homoRefCount),
    alternative_allele_frequency = alt_allele_count / total_allele_count
  ) |>
  select(
    rsid,
    beta = nes,
    se = error,
    pValue,
    minor_allele_frequency = maf,
    alternative_allele_frequency,
    chromosome:genome_build,
    tissueSiteDetailId
  )

With the exception of get_sample_biobank_data() and get_sample_datasets(), for which ‘get_sample’ is additionally appended with their respective category titles ‘biobank_data’ and ‘datasets’.↩︎