Fig. 1: Overview of scPRS and its applications. | Nature Biotechnology

Fig. 1: Overview of scPRS and its applications.

From: Single-cell polygenic risk scores dissect cellular and molecular heterogeneity of complex human diseases

Fig. 1

For a given disease, scPRS first leverages GWAS summary statistics obtained from the discovery cohort and the reference scATAC-seq or snATAC-seq dataset to calculate single-cell-level PRSs with different parameters for individuals in the target cohort. Next, scPRS embeds and propagates cell-level PRSs over the cell–cell similarity network using a GNN. The final readout combines smoothed PRSs from all cells to predict the disease risk. scPRS is trained to minimize the loss between predicted and true disease labels. The trained model can be used to (1) predict disease risk for unseen individuals; (2) prioritize disease-relevant cells and cell types; and (3) fine-map disease risk variants, genes and disrupted genetic regulation in specific cell types. UMAP, uniform manifold approximation and projection. The schematic was created using BioRender.com.

Back to article page