Extended Data Fig. 5: Performance comparison across gene expression imputation methods with per-gene metrics (n=12,557 genes).

(a, b) Per-tissue comparison between HYFA and TEEBoT when using (a) whole-blood and (b) all accessible tissues (whole blood, skin sun exposed, skin not sun exposed, and adipose subcutaneous) as reference. We discarded target tissues represented by less than 25 test individuals. HYFA achieved superior Pearson correlation in (a) 25 out of 48 target tissues when a single tissue was used as reference and (b) all target tissues when multiple reference tissues were considered. For underrepresented target tissues (less than 25 individuals with source and target tissues in the test set), we considered all the validation and test individuals (translucent bars). (c, d) Prediction performance from (c) whole-blood gene expression and (d) accessible tissues as reference. Boxes show quartiles and whiskers depict the distribution range (1.5 times the interquartile range). Mean imputation replaces missing values with the feature averages. Blood surrogate utilises gene expression in whole blood as a proxy for the target tissue. k-Nearest Neighbours (kNN) imputes missing features with the average of measured values across the k nearest observations (k=20). TEEBoT projects reference gene expression into a low- dimensional space with principal component analysis (PCA; 30 components), followed by linear regression to predict target values. HYFA (all) employs information from all collected tissues. Boxes show quartiles, centerlines correspond to the median, and whiskers depict the distribution range (1.5 times the interquartile range). Outliers outside of the whiskers are shown as distinct points.