In maialab/quincunx: REST API Client for the 'PGS' Catalog

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
library(quincunx)

Introduction

Polygenic scores (PGSs) are annotated with information about the phenotype that it predicts, i.e. the reported trait (as reported in the original publication). This can be found as the column reported_trait in slot scores of scores objects:

pgs_01 <- get_scores('PGS000001')
pgs_01@scores

The predicted phenotype is also mapped to Experimental Factor Ontology (EFO) terms (a controlled vocabulary for the unambiguous identification of traits and diseases, and their relationships), namely, the EFO trait. The EFO traits associated with a polygenic score can also be found in scores objects in the slot traits, column trait:

pgs_01@traits

Many PGSs have been developed and demonstrated to be predictive of common complex traits, e.g. body mass index (BMI) [@khera2019], blood lipids [@kuchenbaecker2019] and educational attainment [@lee2018].

Similarly, PGSs for various diseases have been shown to be predictive of disease incidence, defining marked increases in risk over the life course or at earlier ages for people with high PGSs, e.g. coronary artery disease [@inouye2018;@khera2018], breast cancer [@mavaddat2019] and schizophrenia [@zheutlin2019].

Getting catalogued traits from PGS Catalog

If you are interested in retrieving polygenic scores from the Catalog, you might want to search them by the trait they predict. get_scores() is the function that searches for PGSs, however, this function only allows to search by pgs_id, efo_id or pubmed_id. So in order to search by a trait term, we need to first find the associated EFO identifiers (efo_id).

To search for traits (or diseases), you use the function get_traits(). With this function you can search by:

The EFO trait identifier: efo_id;
or by the trait term: a term to be matched in the EFO identifier (efo_id), label, description synonyms, trait categories, or external mapped terms.

The most useful search criteria is the trait term, and that is typically want you will want to use. Unless you already know the EFO trait you are interested in, and are looking for extra details about it, you won't search directly with the EFO identifier.

Basic example

Let's say you are interested in PGSs related to medical condition, stroke. Then you can search for "stroke" with get_traits():

get_traits(trait_term = 'stroke')

As can be seen from the returned traits object, we get a set of six tables (slots) that include several details about stroke.

In the first table traits we got only one row, indicating that this query returned only one trait in the Catalog. This trait is named "stroke" (column trait), and is unambiguously identified by the EFO identifier EFO_0000712.

Exact matching

By default, the trait term is matched exactly. If you want to relax the matching, then indicate with the parameter exact_term set to FALSE. This way you will get, potentially, more results, in this example case, ischemic stroke (HP_0002140) is now also returned:

get_traits(trait_term = 'stroke', exact_term = FALSE)

Subtraits (child traits)

By default, subtraits (child traits), are not retrieved by get_traits(). If you want to get all matching traits and those that are child traits thereof, then indicate with the parameter include_children set to TRUE. Here is an example with "breast cancer":

get_traits(trait_term = 'breast cancer', include_children = TRUE)

The column is_child indicates whether that trait is being retrieved because it is a direct result of the query or not. is_child is TRUE when the trait is returned because it is a child trait of a matching trait, and FALSE if a direct result of the query.

In the case of child traits, the column parent_efo_id indicates the EFO trait identifier of the parent trait, i.e. the direct matching trait, or NA otherwise.