MCC-SP

MCC-SP identifies and ranks causal pathways linking single nucleotide polymorphisms (SNPs) to complex disease phenotypes by combining the maximal correlation coefficient with K shortest paths on gene expression networks.

Key Features:

• Integration of Network Analysis: Leverages gene network structures to capture propagation of genetic variant impacts across interconnected genes and pathways.
• Maximal Correlation Coefficient (MCC): Uses MCC to quantify connection strength between network nodes and, in simulations, outperforms Pearson and Spearman correlations, distance correlation, mutual information, and maximal information coefficient.
• K Shortest Paths Algorithm: Integrates MCC-weighted networks with a K shortest paths algorithm to identify and rank multiple plausible pathway chains from a SNP to a disease phenotype.
• Pathway Importance Score (PIS): Computes a Pathway Importance Score to quantify and prioritize the significance of each identified pathway.
• Cohort application example: Applied to the Religious Orders Study and the Memory and Aging Project to identify two key pathways linking APOE genotype to Alzheimer's disease via gene expression changes relevant to Alzheimer's pathology.

Scientific Applications:

• Mapping regulatory mechanisms of noncoding variants: Systematically maps and ranks causal pathways through which noncoding genetic variants influence gene expression and phenotypes.
• Prioritization of therapeutic targets: Ranks candidate pathways to aid prioritization of targets for downstream functional validation or drug development.
• Cohort-level genotype–phenotype discovery: Enables discovery of genotype→expression→disease links in cohort gene expression datasets such as the Religious Orders Study and the Memory and Aging Project.

Methodology:

Compute maximal correlation coefficients between network nodes to weight gene expression networks, apply a K shortest paths algorithm on the MCC-weighted network to identify and rank pathways, compute a Pathway Importance Score for each pathway, and validate MCC performance via simulations comparing it to Pearson, Spearman, distance correlation, mutual information, and maximal information coefficient.