Clumping and Thresholding

The initial filter that retains only variants with a p-value below a genome-wide significance cutoff, typically 5 × 10⁻⁸. This Bonferroni-corrected threshold accounts for approximately one million independent tests in the human genome. Variants exceeding this threshold are discarded as likely false positives. In practice, researchers may also apply a suggestive threshold (e.g., 1 × 10⁻⁵) to flag regions for follow-up in replication cohorts. The choice of threshold directly controls the family-wise error rate and balances sensitivity against specificity in biomarker discovery pipelines.

After thresholding, clumping iterates through significant variants sorted by p-value. For each index variant, all other variants within a specified LD window (e.g., 250-1000 kilobases) that exceed an r² correlation threshold (typically 0.1-0.5) are pruned. This ensures that each clump represents a single independent association signal rather than multiple correlated proxies. The r² parameter controls clump granularity: lower values produce more clumps by retaining weakly correlated variants, while higher values yield fewer, more stringently independent loci.

An alternative to clumping that uses stepwise model selection to identify multiple independent signals within a locus. The GCTA-COJO method performs:

• Conditional analysis: Tests each variant while conditioning on the top signal
• Joint analysis: Fits all selected variants simultaneously

This approach can resolve allelic heterogeneity where clumping may collapse distinct causal variants in tight LD. It requires summary statistics and a reference LD matrix, making it computationally more intensive but statistically more precise for fine-mapping complex trait loci.

After identifying independent clumps, lead variants are annotated to prioritize functional biomarkers:

• Variant Effect Predictor (VEP): Annotates consequence types (missense, nonsense, regulatory)
• FUMA: Maps clumps to genes via positional, eQTL, and chromatin interaction data
• MAGMA: Performs gene-set enrichment on clumped results

This annotation pipeline bridges statistical association to biological mechanism, enabling researchers to nominate causal genes and pathways for experimental validation in drug target discovery programs.