Direction, not significance: cross-cancer genetic pleiotropy framework, source data and interactive locus map (v1.0.0)
Year: 2026
Type: software
Abstract: Added Related-aware null distribution: trait z-correlation matrix estimated from 634,036 variants, with the global excess of aligned loci recomputed under that null (p1_trait_cor*.csv, src/p1_trait_cor.R). Symmetric selection analysis and locus-level bootstrap (95% intervals for the aligned and reversed counts and for the reversed-arm borrowing fraction) (p1_class_summary.csv, p1_locus_audit.csv). Overlap-corrected generalized least-squares test and common-factor loadings (p1_gls_*.csv, p1_commonfactor_loadings.csv, p1_factor_score.csv). Extended simulation: 42 scenarios covering sample-size ratios down to 1:200 and correlated sampling errors, 10 replicates each (mtag_simulation_extended_raw.csv, src/s12b_mtag_simulation_extended.py). Replication-panel resource: 471 loci x 14 tumours = 6,594 locus-tumour records, with MTAG z, own-GWAS z, borrowing fraction, direction class and significance flags (replication_panel_resource.csv). Changed Interactive locus map and resource files updated to v1.2: z-scores stored to six decimals and the borrowing fraction computed from the stored values, so the fraction reproduces the formula exactly when recomputed from the released file. Notes Direction classes are inferred from MTAG and are descriptive, not tests. Borrowing fractions in the replication panel are semi-quantitative, because every endpoint is defined against one shared control set.
Authors: Wang, Peng, Li, Wen, An, Jianlei, Sun, Guogui
Language: English
Subfield: Genetics
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7219655705