Comparisons of seven algorithms for pathway analysis using the WTCCC Crohn's Disease dataset

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Pathway analysis for genome-wide association studies (GWAS) is increasingly popular for identifying disease mechanisms.
• Existing algorithms vary in input data, hypothesis testing, and analysis stages.
• Permutation strategies are common for evaluating pathway significance due to complex SNP-gene-pathway relationships.

Purpose of the Study:

• To implement and compare two novel GWAS pathway analysis algorithms within the KGG software.
• To evaluate the performance of KGG algorithms against five other selected methods using a real-world GWAS dataset.

Main Methods:

• Two summary statistics-based algorithms were implemented in the KGG tool.
• Comparative analysis was performed on the WTCCC Crohn's Disease dataset using MsigDB canonical pathways.
• Permutation testing was used to calculate empirical p-values for significance.

Main Results:

• Most algorithms controlled Type I error rates, with some being conservative.
• Significant variation was observed in statistical power and computational time among methods.
• PLINK's truncated set-based test demonstrated the highest power, while KGG offered the fastest execution time.

Conclusions:

• Raw data-based algorithms, like those in PLINK, are recommended for GWAS pathway analysis when computational resources permit.
• Utilizing multiple pathway analysis algorithms on the same GWAS data may yield complementary findings for complex diseases.