| _version_ | 1866902171992195072 |
|---|---|
| author | Ashuraliyev, Abduxoliq |
| author_facet | Ashuraliyev, Abduxoliq |
| contents | This repository contains the code, configuration, and reproducibility artifacts for the paper "Path-Probability Models Outperform Point-Estimate Scores for Noncoding GWAS Gene Prioritization". The framework implements probabilistic mechanism graphs with path-probability inference for GWAS gene prioritization, demonstrating that explicit causal paths outperform point-estimate locus-to-gene scores. Key features: - SuSiE-based fine-mapping with multi-signal support - coloc.susie colocalization for multi-causal variant handling - ABC (Activity-by-Contact) and PCHi-C enhancer-gene linking - Noisy-OR probabilistic inference model - Per-module probability calibration (ECE < 0.05) - eQTL Catalogue cross-study replication (78% replication rate) - Three-tier benchmark system with anti-leakage provisions Results: - 76% recall at rank 20 vs 58% for Open Targets L2G - Effect size correlation r=0.89 with eQTL Catalogue - Calibrated probabilities at each module Full pipeline reproducible via Snakemake workflow. |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17799514 |
| institution | Zenodo |
| language | eng |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Mechanism-First Causal Graphs for Noncoding GWAS: Code and Data Ashuraliyev, Abduxoliq GWAS fine-mapping colocalization causal graphs gene prioritization eQTL regulatory genomics probabilistic inference mechanism graphs SuSiE ABC model PCHi-C coloc.susie path-probability This repository contains the code, configuration, and reproducibility artifacts for the paper "Path-Probability Models Outperform Point-Estimate Scores for Noncoding GWAS Gene Prioritization". The framework implements probabilistic mechanism graphs with path-probability inference for GWAS gene prioritization, demonstrating that explicit causal paths outperform point-estimate locus-to-gene scores. Key features: - SuSiE-based fine-mapping with multi-signal support - coloc.susie colocalization for multi-causal variant handling - ABC (Activity-by-Contact) and PCHi-C enhancer-gene linking - Noisy-OR probabilistic inference model - Per-module probability calibration (ECE < 0.05) - eQTL Catalogue cross-study replication (78% replication rate) - Three-tier benchmark system with anti-leakage provisions Results: - 76% recall at rank 20 vs 58% for Open Targets L2G - Effect size correlation r=0.89 with eQTL Catalogue - Calibrated probabilities at each module Full pipeline reproducible via Snakemake workflow. |
| title | Mechanism-First Causal Graphs for Noncoding GWAS: Code and Data |
| topic | GWAS fine-mapping colocalization causal graphs gene prioritization eQTL regulatory genomics probabilistic inference mechanism graphs SuSiE ABC model PCHi-C coloc.susie path-probability |
| url | https://doi.org/10.5281/zenodo.17799514 |