A supervised approach for identifying discriminating genotype patterns and its application to breast cancer data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Large-scale genetic studies generate vast genomic variation data.
• Identifying discriminating genotypic patterns is challenging due to numerous single nucleotide polymorphism (SNP) combinations.
• Integrating additional high-throughput data, like gene expression, complicates pattern detection.

Purpose of the Study:

• To develop a graph theoretic approach for identifying discriminating patterns (DPs) for phenotypes.
• To enable the detection of genotype-phenotype associations in large human cohorts.
• To integrate and analyze multi-omics data, including SNP and gene expression profiles.

Main Methods:

• Representing SNP data as a bipartite graph of individuals and their SNP states.
• Identifying fully connected subgraphs that link individuals with specific phenotypic groups.
• Employing a supervised, phenotype-guided search process, distinct from traditional biclustering.

Main Results:

• The method successfully retrieved planted patterns in simulations.
• Applied to breast cancer data, it detected significant DPs associated with clinical phenotypes.
• Identified patient groups enriched for other phenotypes and exhibiting expression coherency.
• Detected gene groups with functional coherency, including genes with known roles in cancer.

Conclusions:

• The graph theoretic approach effectively identifies phenotype-discriminating genetic patterns.
• The method integrates genomic and gene expression data for robust association studies.
• Detected patterns and associated gene sets provide insights into cancer biology.