Exhaustive Variant Interaction Analysis using Multifactor Dimensionality Reduction.

Area of Science:

• Human Genetics
• Computational Biology
• Genomics

Background:

• Understanding genomic variation's link to complex disorders like Type 2 Diabetes (T2D) is crucial.
• Traditional methods often overlook interactions between genomic variants, which may significantly influence disease development.
• Studying these interactions is computationally challenging but essential for a complete understanding of complex disease genetics.

Approach:

• Leveraged High-Performance Computing (HPC) and machine learning methods.
• Developed a containerized framework utilizing Multifactor Dimensionality Reduction (MDR).
• Applied the framework to a large dataset from the Northwestern University NUgene project cohort.

Key Points:

• Analyzed 1,883,192 variant pairs for association with Type 2 Diabetes (T2D).
• Identified 104 significant variant pairs associated with T2D.
• Discovered two variant pairs with potential functional relevance to T2D.

Conclusions:

• High-Performance Computing and machine learning effectively address the computational demands of studying gene-gene interactions.
• The developed framework successfully identified significant genomic variant pairs linked to Type 2 Diabetes.
• Further investigation into the functional roles of identified variant pairs could enhance our understanding of T2D etiology.