Biological networks provide a natural framework for integrating the diverse genetic variations associated with complex and multifactorial disorders. The main challenge in the analysis of rare genetic variations, such as de novo single nucleotide polymorphisms (SNPs) and copy number variations (CNVs) — duplications or deletions of stretches of DNA — is precisely their rarity. With currently available sample sizes, a vast majority of the observed genetic events are either unique or show only limited recurrency.
To identify affected molecular networks, Dennis Vitkup and his colleagues at Columbia University have developed a computational approach called NETBAG that searches for cohesive clusters of genes perturbed by disease-associated genetic variants. The approach is based on the underlying phenotype network, which assigns every pair of human genes a score that is proportional to the likelihood that these genes are involved in the same genetic phenotype.
NETBAG integrates data from multiple types of genetic variation: SNPs, CNVs and loci implicated by genome-wide association studies. The NETBAG search algorithm identifies highly connected gene clusters that are affected by genetic variants, then calculates the statistical significance of the identified clusters using randomized input data.
The researchers’ network-based analysis of genes harboring de novo mutations in individuals with autism suggests that affected genes converge on several key biological processes: the postsynaptic density, cell-cell interactions and mobility, channel and receptor activities, chromatin modification and regulation, intercellular signaling and the cytoskeleton. Each of these broad processes contains multiple contributing pathways that are involved in synaptic and neuronal functions across brain regions and during various stages of brain development.
Vitkup and his team’s analyses demonstrate that functional properties of implicated genes affect the phenotypes of the disorder. Stronger functional impact, as determined by mutation types, brain expression or network properties, leads to more severely affected disease phenotypes that exhibit intellectual, social or behavioral abnormalities.