A Novel Approach to Detecting Epistasis using Random Sampling Regularisation.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Epistasis, the interaction of multiple genetic variants, complements the 'common disease, common variant' hypothesis.
• Traditional epistasis analysis faces challenges with high-order interactions, increasing the False Discovery Rate (FDR) and computational complexity.
• Existing methods struggle with the factorial time complexity (O(n!)) of multivariate epistasis analysis.

Purpose of the Study:

• To propose a novel methodology for detecting epistasis using interpretable methods.
• To outline genetic interactions through advanced filtering processes.
• To address the limitations of computational complexity and FDR in epistasis analysis.

Main Methods:

• Developed a new methodology for epistasis detection.
• Employed Random Sampling Regularisation to create sample sets for a voting system.
• Utilized filtering processes to identify significant biological markers (SNPs) and their interactions.

Main Results:

• Preliminary results show promising and concise detection of genetic interactions.
• Identified eight risk candidate interactions involving five variants in breast cancer classification.
• Discovered a single candidate variant with a high protective association.

Conclusions:

• The novel methodology offers a more reliable and computationally feasible approach to epistasis detection.
• This method can effectively identify complex genetic interactions relevant to disease classification, such as in breast cancer.
• The findings pave the way for better understanding of genetic contributions to complex diseases.