compute

PolyGenius data transformation engine

Description

compute provides a high-level interface for data transformations on PolyGeniusData objects. It organizes functions for:

• compute$scores(): Calculate PRS scores from genotypes

• compute$populationStructure(): Compute principal components for population stratification correction

• compute$kinship(): Calculate relatedness matrix

• compute$singleVariantAssociations(): Perform GWAS on PRS variants

• compute$similarity$obs(): Compute observation-observation similarity matrices

• compute$similarity$mod(): Compute model-model similarity matrices

• compute$embedding$obs(): Compute low-dimensional embeddings for observations

• compute$embedding$mod(): Compute low-dimensional embeddings for models

Provides a compact overview of the data transformation interface.

Usage

compute

print.computeEngine(x, ...)

Arguments

x	An object of class “computeEngine”, typically the exported compute environment.
…	Unused. Present for S3 compatibility.

Format

An environment with components:

• scores(): wrapper for compute.scores()

• populationStructure(): wrapper for compute.population.structure()

• kinship(): wrapper for compute.kinship()

• singleVariantAssociations(): wrapper for compute.single.variant.associations()

• similarity: environment with obs() and mod() functions

• embedding: environment with obs() and mod() functions

Details

A typical compute workflow:

1. Calculate PRS scores:

   compute$scores(data, maf.threshold = 0.01)
   

2. Compute population structure:

   compute$populationStructure(data,
                               npcs = 10,
                               reference.panel = "EUR")
   

3. Calculate kinship matrix:

   compute$kinship(data)
   

4. Compute similarity and embeddings:

   # Observation similarity
   sim.obs <- compute$similarity$obs(data, scores, method = "correlation")
   # Model similarity
   sim.mod <- compute$similarity$mod(data, mod, method = "snp.jaccard")
   # PCA embedding
   pca.emb <- compute$embedding$obs(data, scores, method = "pca", n.components = 10)
   # UMAP embedding
   umap.emb <- compute$embedding$obs(data, scores, method = "umap")
   

5. Run single-variant associations:

   compute$singleVariantAssociations(data,
                                     phenotypes = c(case_status, age_onset),
                                     control.for = c(age, sex, PC1_10))
   

Value

The input object x, invisibly.