Start by loading {gwasrapidd}
, and
{dplyr}
and {tidyr}
:
Let’s say you want to retrieve all variants associated with the phenotype Body Mass Index (BMI). Moreover, you want to sort them by their risk allele (minor allele), as well as the effect size (beta coefficient) and p-value.
First we start by finding the Experimental Factor Ontology (EFO) identifier(s) corresponding to BMI. To do this, we start by downloading all traits in the GWAS Catalog.
Then look for ‘BMI’ in the trait description column.
dplyr::filter(all_traits@traits, grepl('BMI', trait, ignore.case = TRUE))
#> # A tibble: 10 × 3
#> efo_id trait uri
#> <chr> <chr> <chr>
#> 1 EFO_0005937 longitudinal BMI measurement http://www.ebi.ac.uk/efo/EFO_0005937
#> 2 EFO_0007737 BMI-adjusted adiponectin measurement http://www.ebi.ac.uk/efo/EFO_0007737
#> 3 EFO_0007788 BMI-adjusted waist-hip ratio http://www.ebi.ac.uk/efo/EFO_0007788
#> 4 EFO_0007789 BMI-adjusted waist circumference http://www.ebi.ac.uk/efo/EFO_0007789
#> 5 EFO_0007793 BMI-adjusted leptin measurement http://www.ebi.ac.uk/efo/EFO_0007793
#> 6 EFO_0008036 BMI-adjusted fasting blood glucose measurement http://www.ebi.ac.uk/efo/EFO_0008036
#> 7 EFO_0008037 BMI-adjusted fasting blood insulin measurement http://www.ebi.ac.uk/efo/EFO_0008037
#> 8 EFO_0008038 BMI-adjusted hip bone size http://www.ebi.ac.uk/efo/EFO_0008038
#> 9 EFO_0008039 BMI-adjusted hip circumference http://www.ebi.ac.uk/efo/EFO_0008039
#> 10 EFO_0011044 BMI-adjusted neck circumference http://www.ebi.ac.uk/efo/EFO_0011044
So there are several phenotypes whose description includes the
keyword ‘BMI’. However, only the EFO trait 'EFO_0005937'
(‘longitudinal BMI measurement’) really corresponds to BMI as a
phenotypic trait. All other traits are adjusted for BMI but are not BMI
traits per se (you can further confirm this by looking at each trait
description, just by opening your web browser with each respective
URI).
To get statistical association data for the trait ‘longitudinal BMI
measurement’ ('EFO_0005937'
), as well as associated
variants and effect sizes, we use the gwasrapidd
get_associations()
function:
The S4 object bmi_associations
contains several tables,
namely 'associations'
, 'loci'
,
'risk_alleles'
, 'genes'
,
'ensembl_ids'
and 'entrez_ids'
:
slotNames(bmi_associations)
#> [1] "associations" "loci" "risk_alleles" "genes" "ensembl_ids" "entrez_ids"
From table 'associations'
we can extract the
variables:
'association_id'
'pvalue'
'beta_number'
'beta_unit'
'beta_direction'
whereas from table 'risk_alleles'
we can obtain:
'association_id'
'variant_id'
'risk_allele'
.We extract all these variables and combine them into one single
dataframe (bmi_variants
), using
'association_id'
as the matching key:
tbl01 <- dplyr::select(bmi_associations@risk_alleles, association_id, variant_id, risk_allele)
tbl02 <- dplyr::select(bmi_associations@associations, association_id, pvalue, beta_number, beta_unit, beta_direction)
bmi_variants <- dplyr::left_join(tbl01, tbl02, by = 'association_id') %>%
tidyr::drop_na() %>%
dplyr::arrange(variant_id, risk_allele)
The final results show 42 associations. Note that some variant/allele combinations might be repeated as the same variant/allele combination might have been assessed in more than one GWAS study.
bmi_variants
#> # A tibble: 42 × 7
#> association_id variant_id risk_allele pvalue beta_number beta_unit beta_direction
#> <chr> <chr> <chr> <dbl> <dbl> <chr> <chr>
#> 1 10066323 rs10041997 A 0.0000004 0.138 unit increase
#> 2 61729722 rs10070777 A 0.000004 0.879 unit increase
#> 3 61729714 rs10278819 C 0.000002 0.873 unit increase
#> 4 61729787 rs10426669 G 0.000006 0.765 unit increase
#> 5 61729731 rs1048163 G 0.000005 0.450 unit increase
#> 6 61729771 rs1048164 A 0.000009 0.436 unit increase
#> 7 55309232 rs10515235 A 0.000002 0.05 unit increase
#> 8 55309249 rs10938397 G 0.00000003 0.06 unit increase
#> 9 61729763 rs112045010 C 0.000005 0.906 unit increase
#> 10 61729759 rs11979775 C 0.000005 0.501 unit increase
#> # … with 32 more rows