Interpretation Boundary
This workflow is designed for candidate prioritization and hypothesis generation. It does not convert model scores into clinical risk probabilities or validated functional effects.
NSOFC-RegRank candidate variant workflow
Workflow record for GWAS-derived candidate variant organization, coordinate harmonization, allele auditing, regulatory annotation, and model-score coverage checks.
Reproducibility
Repository status
Complete reproduction materials are available but not yet public. A GitHub repository will be linked here when it is ready.
Environment
Language: Python / R / shell Package: project-specific OS: macOS / Linux Hardware: CPU workflow for most data-auditing steps
Reproduction Steps
- Collect GWAS source records and normalize phenotype, coordinate, rsID, REF/ALT, and association-direction fields.
- Define LD-expanded candidate variant spaces using ancestry-matched LD references.
- Build craniofacial regulatory annotation features after coordinate harmonization and quality control.
- Audit sparse positive labels separately from regulatory-overlap annotations to avoid label leakage.
- Compare transparent baselines and external model scores only within documented model-coverage boundaries.
Expected Outputs
- Candidate variant tables with traceable coordinate and allele fields.
- Functional benchmark labels with strict positive, silver-standard positive, and unlabeled candidate categories.
- Coverage-aware model comparison tables and conservative candidate-prioritization summaries.
Next steps
Explore related project material
Previous HERMES-IC+ feature and evaluation workflow
Next Latest entry in this sequence