PGS Construction Process and Stages

Overview of polygenic risk score construction

PGS is calculated as a weighted sum of several risk variants from a genome-wide association study in one or more samples with multiple p-value thresholds. The effect sizes are typically estimated as \(\beta\) (beta coefficients) or as odds ratios. After the PGS is calculated in one sample, the distribution of individual PGS is assessed in another in an independent sample set.

Adapted from Konuma & Okada (2021), CC BY 4.0.

PGS development and evaluation stages

In the PGS Catalog, cohorts and samples are annotated according to their utilisation context, i.e. stage, in the PGS construction process. In quincunx, the stage is indicated by the stage variable that can have one of these values:

• gwas: to annotate samples used to derive variant associations (GWAS)
• dev: to annotate samples used in the development or training of PGSs
• gwas/dev: as a catch-all term to annotate samples used either in gwas or dev stages
• eval: to annotate samples used in the PGS evaluation stage

You will encounter the stage annotation in tables of objects returned by quincunx’s retrieval functions. Here are a few examples:

In a scores object

get_scores('PGS000327')@samples
#> # A tibble: 2 × 15
#>   pgs_id   sampl…¹ stage sampl…² sampl…³ sampl…⁴ sampl…⁵ pheno…⁶ ances…⁷ ances…⁸
#>   <chr>      <int> <chr>   <int>   <int>   <int>   <dbl> <chr>   <chr>   <chr>  
#> 1 PGS0003…       1 gwas    46350      NA      NA      NA Cases … Europe… NA     
#> 2 PGS0003…       2 dev     28592   10461   18131      NA Cases … Europe… NA     
#> # … with 5 more variables: country <chr>,
#> #   ancestry_additional_description <chr>, study_id <chr>, pubmed_id <chr>,
#> #   cohorts_additional_description <chr>, and abbreviated variable names
#> #   ¹​sample_id, ²​sample_size, ³​sample_cases, ⁴​sample_controls,
#> #   ⁵​sample_percent_male, ⁶​phenotype_description, ⁷​ancestry_category, ⁸​ancestry
#> # ℹ Use `colnames()` to see all variable names

In a sample_sets object

get_sample_sets(pgs_id = 'PGS000327')@samples
#> # A tibble: 1 × 15
#>   pss_id   sampl…¹ stage sampl…² sampl…³ sampl…⁴ sampl…⁵ pheno…⁶ ances…⁷ ances…⁸
#>   <chr>      <int> <chr>   <int>   <int>   <int>   <dbl> <chr>   <chr>   <chr>  
#> 1 PSS0004…       1 eval     7148    2615    4532      NA Cases … Europe… NA     
#> # … with 5 more variables: country <chr>,
#> #   ancestry_additional_description <chr>, study_id <chr>, pubmed_id <chr>,
#> #   cohorts_additional_description <chr>, and abbreviated variable names
#> #   ¹​sample_id, ²​sample_size, ³​sample_cases, ⁴​sample_controls,
#> #   ⁵​sample_percent_male, ⁶​phenotype_description, ⁷​ancestry_category, ⁸​ancestry
#> # ℹ Use `colnames()` to see all variable names

In a performance_metrics object

get_performance_metrics(pgs_id = 'PGS000327')@samples
#> # A tibble: 1 × 16
#>   ppm_id    pss_id sampl…¹ stage sampl…² sampl…³ sampl…⁴ sampl…⁵ pheno…⁶ ances…⁷
#>   <chr>     <chr>    <int> <chr>   <int>   <int>   <int>   <dbl> <chr>   <chr>  
#> 1 PPM000879 PSS00…       1 eval     7148    2615    4532      NA Cases … Europe…
#> # … with 6 more variables: ancestry <chr>, country <chr>,
#> #   ancestry_additional_description <chr>, study_id <chr>, pubmed_id <chr>,
#> #   cohorts_additional_description <chr>, and abbreviated variable names
#> #   ¹​sample_id, ²​sample_size, ³​sample_cases, ⁴​sample_controls,
#> #   ⁵​sample_percent_male, ⁶​phenotype_description, ⁷​ancestry_category
#> # ℹ Use `colnames()` to see all variable names

In the stages_tally table:

get_scores('PGS000327')@stages_tally
#> # A tibble: 3 × 4
#>   pgs_id    stage sample_size n_sample_sets
#>   <chr>     <chr>       <int>         <int>
#> 1 PGS000327 gwas        46350            NA
#> 2 PGS000327 dev         28592            NA
#> 3 PGS000327 eval           NA             1

In the ancestry_frequencies table:

get_scores('PGS000012')@ancestry_frequencies
#> # A tibble: 4 × 4
#>   pgs_id    stage ancestry_class_symbol frequency
#>   <chr>     <chr> <chr>                     <dbl>
#> 1 PGS000012 gwas  MAE                         100
#> 2 PGS000012 dev   EUR                         100
#> 3 PGS000012 eval  EUR                          75
#> 4 PGS000012 eval  MAE                          25

And in multi_ancestry_composition table:

get_scores('PGS000012')@multi_ancestry_composition
#> # A tibble: 4 × 4
#>   pgs_id    stage multi_ancestry_class_symbol ancestry_class_symbol
#>   <chr>     <chr> <chr>                       <chr>                
#> 1 PGS000012 gwas  MAE                         EUR                  
#> 2 PGS000012 gwas  MAE                         SAS                  
#> 3 PGS000012 eval  MAE                         EUR                  
#> 4 PGS000012 eval  MAE                         NR

In a cohorts object:

get_cohorts('23andMe')@pgs_ids
#> # A tibble: 18 × 3
#>    cohort_symbol pgs_id    stage   
#>    <chr>         <chr>     <chr>   
#>  1 23andMe       PGS000336 gwas/dev
#>  2 23andMe       PGS000730 gwas/dev
#>  3 23andMe       PGS000731 gwas/dev
#>  4 23andMe       PGS000732 gwas/dev
#>  5 23andMe       PGS000766 gwas/dev
#>  6 23andMe       PGS000767 gwas/dev
#>  7 23andMe       PGS000780 gwas/dev
#>  8 23andMe       PGS000790 gwas/dev
#>  9 23andMe       PGS000800 gwas/dev
#> 10 23andMe       PGS000810 gwas/dev
#> 11 23andMe       PGS000867 gwas/dev
#> 12 23andMe       PGS000902 gwas/dev
#> 13 23andMe       PGS000903 gwas/dev
#> 14 23andMe       PGS001774 gwas/dev
#> 15 23andMe       PGS002235 gwas/dev
#> 16 23andMe       PGS000730 eval    
#> 17 23andMe       PGS000731 eval    
#> 18 23andMe       PGS000732 eval