Extended Data Fig. 3: Top predicted Alzheimer’s disease-relevant genes in multiple brain regions, with whole blood as reference tissue.
From: Hypergraph factorization for multi-tissue gene expression imputation

(a) Pearson correlation coefficient of top 20 predicted genes from the Alzheimer’s disease pathway (KEGG), ranked by average correlation. (b, c, d) Average per-gene expression (x-axis) versus prediction performance (Pearson correlation between predicted and ground truth expression) in (b) cerebellum, (c) cortex, and (d) hippocampus. HYFA exhibits strong prediction performance for several Alzheimer’s disease-relevant genes including APOE (cortex ρ=0.536, cerebellum: ρ=0.502), APP (cortex ρ=0.524), PSEN1 (cerebellum: ρ=0.459), and PSEN2 (cortex: ρ=0.590, cerebellum: ρ=0.559, hippocampus: ρ=0.403). In cerebellum, PSEN1 (ρ=0.459), PSEN2 (ρ=0.559), and APOE (ρ=0.502) attained above expected performances (average ρ=0.448). APP (ρ=0.524), PSEN2 (ρ=0.590), and APOE (ρ=0.536) surpassed the expected correlation in cortex (average ρ=0.443).