Skip to main content
Fig. 1 | Genome Medicine

Fig. 1

From: Genetic feature engineering enables characterisation of shared risk factors in immune-mediated diseases

Fig. 1

Schematic of basis creation and projection. Basis creation: GWAS summary statistics for related traits are combined to create a matrix, M (n × m), of harmonised effect sizes (\( \hat{\beta} \)) and a learnt vector of shrinkage values for each SNP. After multiplying each row of M by the shrinkage vector, PCA is used to decompose M into component and loading matrices. Basis projection: for an independent set of studies, trait effects are harmonised with respect to the basis, shrinkage applied, and the resultant vector is multiplied by the basis loading matrix to obtain component scores. These component scores can be used for testing hypotheses of the form that a weighted average of effect sizes in the test GWAS is non-zero, because the weights (basis loading matrix) are learnt from an independent set of large GWAS

Back to article page