TuRF-E

TuRF-E applies an ensemble filtering approach to stabilize and improve detection of single nucleotide polymorphism (SNP) interactions in genome-wide association (GWA) studies.

Key Features:

• Ensemble filtering: Aggregates results from multiple runs of ReliefF and TuRF (tuned ReliefF) using ensemble learning principles to produce consolidated SNP rankings.
• Robustness to sample order: Mitigates sensitivity to sample order by varying sample order across runs and combining outcomes to yield more consistent SNP rankings.
• Variants ReliefF-E and TuRF-E: Implements ensemble extensions named ReliefF-E and TuRF-E that apply the aggregation strategy to ReliefF and tuned ReliefF respectively.
• Improved retention of interactions: Enhances retention of significant gene-gene interactions relative to conventional metrics such as χ²-tests and odds ratio calculations by synthesizing multiple algorithmic runs.

Scientific Applications:

• Genome-wide association studies (GWA): Filtering SNP interactions in large-scale GWA datasets containing millions of SNPs.
• Complex disease genetics: Prioritizing candidate gene-gene interactions involved in complex disease etiology.
• Large-scale SNP analysis: Narrowing candidate genetic contributors from vast SNP datasets for downstream interaction analysis.

Methodology:

Run ReliefF and TuRF (tuned ReliefF) multiple times with varied sample orders and aggregate the run results using ensemble learning principles to stabilize SNP rankings.