ABSTRACT
Most heritable diseases are polygenic. To comprehend the underlying genetic architecture, it is crucial to discover the clinically relevant epistatic interactions (EIs) between genomic single nucleotide polymorphisms (SNPs) (1-3). Existing statistical computational methods for EI detection are mostly limited to pairs of SNPs due to the combinatorial explosion of higher-order EIs. With NeEDL (network-based epistasis detection via local search), we leverage network medicine to inform the selection of EIs that are an order of magnitude more statistically significant compared to existing tools and consist, on average, of five SNPs. We further show that this computationally demanding task can be substantially accelerated once quantum computing hardware becomes available. We apply NeEDL to eight different diseases and discover genes (affected by EIs of SNPs) that are partly known to affect the disease, additionally, these results are reproducible across independent cohorts. EIs for these eight diseases can be interactively explored in the Epistasis Disease Atlas (https://epistasis-disease-atlas.com). In summary, NeEDL demonstrates the potential of seamlessly integrated quantum computing techniques to accelerate biomedical research. Our network medicine approach detects higher-order EIs with unprecedented statistical and biological evidence, yielding unique insights into polygenic diseases and providing a basis for the development of improved risk scores and combination therapies.
ABSTRACT
BACKGROUND: Asthma with severe exacerbation is one of the most common causes of hospitalization among young children. Exacerbations are typically triggered by respiratory infections, but the host factors causing recurrent infections and exacerbations in some children are poorly understood. As a result, current treatment options and preventive measures are inadequate. OBJECTIVE: We sought to identify genetic interaction associated with the development of childhood asthma. METHODS: We performed an exhaustive search for pairwise interaction between genetic single nucleotide polymorphisms using 1204 cases of a specific phenotype of early childhood asthma with severe exacerbations in patients aged 2 to 6 years combined with 5328 nonasthmatic controls. Replication was attempted in 3 independent populations, and potential underlying immune mechanisms were investigated in the COPSAC2010 and COPSAC2000 birth cohorts. RESULTS: We found evidence of interaction, including replication in independent populations, between the known childhood asthma loci CDHR3 and GSDMB. The effect of CDHR3 was dependent on the GSDMB genotype, and this interaction was more pronounced for severe and early onset of disease. Blood immune analyses suggested a mechanism related to increased IL-17A production after viral stimulation. CONCLUSIONS: We found evidence of interaction between CDHR3 and GSDMB in development of early childhood asthma, possibly related to increased IL-17A response to viral infections. This study demonstrates the importance of focusing on specific disease subtypes for understanding the genetic mechanisms of asthma.
Subject(s)
Asthma , Genome-Wide Association Study , Asthma/genetics , Cadherin Related Proteins , Cadherins/genetics , Genetic Predisposition to Disease , Humans , Interleukin-17/genetics , Membrane Proteins/genetics , Neoplasm Proteins/genetics , Polymorphism, Single Nucleotide , Pore Forming Cytotoxic ProteinsABSTRACT
Background: Recent research reported height biased migration of taller individuals and a Monte Carlo simulation showed that such preferential migration of taller individuals into network hubs can induce a secular trend of height. In the simulation model taller agents in the hubs raise the overall height of all individuals in the network by a community effect. However, it could be seen that the actual network structure influences the strength of this effect. In this paper the background and the influence of the network structure on the strength of the secular trend by migration is investigated. Material and methods: Three principal network types are analyzed: networks derived from street connections in Switzerland, more regular fishing net like networks and randomly generated ones. Our networks have between 10 and 152 nodes and between 20 and 307 edges connecting the nodes. Depending on the network size between 5.000 and 90.000 agents with an average height of 170cm (SD 6.5cm) are initially released into the network. In each iteration new agents are regenerated based on the actual average body height of the previous iteration and, to a certain proportion, corrected by body heights in the neighboring nodes. After generating new agents, a certain number of them migrated into neighbor nodes, the model let preferentially taller agents migrate into network hubs. Migration is balanced by back migration of the same number of agents from nodes with high centrality measures to less connected nodes. The latter is random as well, but not biased by the agents height. Furthermore the distribution of agents per node and their correlation to the centrality of the nodes is varied in a systematic manner. After 100 iterations, the secular trend, i.e. the gain in body height for the different networks, is investigated in relation to the network properties. Results: We observe an increase of average agent body height after 100 iterations if height biased migration is enabled. The increase rate depends on the height of the neighboring factor, the population distribution, the relationship between population in the nodes and their centrality as well as on the network topology. Networks with uniform like distributions of the agents in the nodes, uncorrelated associations between node centrality and agent number per node, as well as very heterogeneous networks with very different node centralities lead to biggest gains in average body height. Conclusion: Our simulations show, that height biased migration into network hubs can possibly contribute to the secular trend of height increase in the human population. The strength of this "tall by migration" event depends on the actual properties of the underlying network. There is a possible significance of this mechanism for social networks, when hubs are represented by individuals and edges as their personal relationships. However, the required high number of iterations to achieve significant effects in more natural network structures in our models requires further studies to test the relevance and real effect sizes in real world scenarios.