Limits...
Mutations Causing Complex Disease May under Certain Circumstances Be Protective in an Epidemiological Sense.

Siegert S, Wolf A, Cooper DN, Krawczak M, Nothnagel M - PLoS ONE (2015)

Bottom Line: In conclusion, a substantial proportion of mutations causing common complex diseases may appear 'protective' in genetic epidemiological studies and hence would normally tend to be excluded, albeit erroneously, from further study.This apparently paradoxical result is explicable in terms of mutual confounding of the respective genotype-phenotype relationships due to a negative correlation between causal mutations induced by their common gene genealogy.As would be predicted by our findings, a significant negative correlation became apparent in published genome-wide association studies between the OR of genetic variants associated with a particular disease and the prevalence of that disease.

View Article: PubMed Central - PubMed

Affiliation: Cologne Center for Genomics, University of Cologne, Cologne, Germany; Institute of Epidemiology, Christian-Albrechts University, Kiel, Germany.

ABSTRACT
Guided by the practice of classical epidemiology, research into the genetic basis of complex disease has usually taken for granted the dictum that causative mutations are invariably over-represented among clinically affected as compared to unaffected individuals. However, we show that this supposition is not true and that a mutation contributing to the etiology of a complex disease can, under certain circumstances, be depleted among patients. Populations with defined disease prevalence were repeatedly simulated under a Wright-Fisher model, assuming various types of population history and genotype-phenotype relationship. For each simulation, the resulting mutation-specific population frequencies and odds ratios (ORs) were evaluated. In addition, the relationship between mutation frequency and OR was studied using real data from the NIH GWAS catalogue of reported phenotype associations of single-nucleotide polymorphisms (SNPs). While rare diseases (prevalence <1%) were found to be consistently caused by rare mutations with ORs>1, up to 20% of mutations causing a pandemic disease (prevalence 10-20%) had ORs<1, and their population frequency ranged from 0% to 100%. Moreover, simulation-based ORs exhibited a wide distribution, irrespective of mutation frequency. In conclusion, a substantial proportion of mutations causing common complex diseases may appear 'protective' in genetic epidemiological studies and hence would normally tend to be excluded, albeit erroneously, from further study. This apparently paradoxical result is explicable in terms of mutual confounding of the respective genotype-phenotype relationships due to a negative correlation between causal mutations induced by their common gene genealogy. As would be predicted by our findings, a significant negative correlation became apparent in published genome-wide association studies between the OR of genetic variants associated with a particular disease and the prevalence of that disease.

No MeSH data available.


Related in: MedlinePlus

Observed proportion of epidemiologically protective causative mutations vs. number of unlinked loci underlying disease etiology.Bold solid line: pandemic disease (prevalence 10%-20%); Thin solid line: common disease (1%-5%); Dashed line: rare disease (0.1%-1.0%). (A) Multiplicative model with parameter value γ = 0.3 (blue) or γ = 0.1 (red). (B) Logistic model with parameter values α = -5; β = 1 (blue) or α = -5; β = 0.5 (red).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4498598&req=5

pone.0132150.g007: Observed proportion of epidemiologically protective causative mutations vs. number of unlinked loci underlying disease etiology.Bold solid line: pandemic disease (prevalence 10%-20%); Thin solid line: common disease (1%-5%); Dashed line: rare disease (0.1%-1.0%). (A) Multiplicative model with parameter value γ = 0.3 (blue) or γ = 0.1 (red). (B) Logistic model with parameter values α = -5; β = 1 (blue) or α = -5; β = 0.5 (red).

Mentions: Analyses of pairs of unlinked (i.e. freely recombining) loci yielded results similar to those of the single-locus analysis (Figs 5 and 6). In fact, the median number of mutations per population, per case and per control remained virtually unchanged under both penetrance models and in all three prevalence categories (Table 3). However, the proportion of mutations with OR<1 was substantially reduced, particularly in instances of high prevalence and MAF (Table 4). This notwithstanding, the simulations revealed that, even with two unlinked loci contributing to the etiology of a given pandemic disease, up to 20% of mutations of at least moderate epidemiological effect size (OR>1.5 or OR<1/1.5) may still appear “protective” (Table 4). Simulations involving 5 or 10 unlinked loci (S1, S2, S3, S4 Figs and S3, S4, S5, S6 Tables) yielded smaller albeit still substantial proportions of apparently protective mutations. Simultaneous consideration of larger numbers of loci unfortunately turned out to be computationally prohibitive. However, the emerging trend (Fig 7) indicates that even polygenic diseases may feature a non-negligible proportion of causative mutations that appear protective in an epidemiological sense.


Mutations Causing Complex Disease May under Certain Circumstances Be Protective in an Epidemiological Sense.

Siegert S, Wolf A, Cooper DN, Krawczak M, Nothnagel M - PLoS ONE (2015)

Observed proportion of epidemiologically protective causative mutations vs. number of unlinked loci underlying disease etiology.Bold solid line: pandemic disease (prevalence 10%-20%); Thin solid line: common disease (1%-5%); Dashed line: rare disease (0.1%-1.0%). (A) Multiplicative model with parameter value γ = 0.3 (blue) or γ = 0.1 (red). (B) Logistic model with parameter values α = -5; β = 1 (blue) or α = -5; β = 0.5 (red).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4498598&req=5

pone.0132150.g007: Observed proportion of epidemiologically protective causative mutations vs. number of unlinked loci underlying disease etiology.Bold solid line: pandemic disease (prevalence 10%-20%); Thin solid line: common disease (1%-5%); Dashed line: rare disease (0.1%-1.0%). (A) Multiplicative model with parameter value γ = 0.3 (blue) or γ = 0.1 (red). (B) Logistic model with parameter values α = -5; β = 1 (blue) or α = -5; β = 0.5 (red).
Mentions: Analyses of pairs of unlinked (i.e. freely recombining) loci yielded results similar to those of the single-locus analysis (Figs 5 and 6). In fact, the median number of mutations per population, per case and per control remained virtually unchanged under both penetrance models and in all three prevalence categories (Table 3). However, the proportion of mutations with OR<1 was substantially reduced, particularly in instances of high prevalence and MAF (Table 4). This notwithstanding, the simulations revealed that, even with two unlinked loci contributing to the etiology of a given pandemic disease, up to 20% of mutations of at least moderate epidemiological effect size (OR>1.5 or OR<1/1.5) may still appear “protective” (Table 4). Simulations involving 5 or 10 unlinked loci (S1, S2, S3, S4 Figs and S3, S4, S5, S6 Tables) yielded smaller albeit still substantial proportions of apparently protective mutations. Simultaneous consideration of larger numbers of loci unfortunately turned out to be computationally prohibitive. However, the emerging trend (Fig 7) indicates that even polygenic diseases may feature a non-negligible proportion of causative mutations that appear protective in an epidemiological sense.

Bottom Line: In conclusion, a substantial proportion of mutations causing common complex diseases may appear 'protective' in genetic epidemiological studies and hence would normally tend to be excluded, albeit erroneously, from further study.This apparently paradoxical result is explicable in terms of mutual confounding of the respective genotype-phenotype relationships due to a negative correlation between causal mutations induced by their common gene genealogy.As would be predicted by our findings, a significant negative correlation became apparent in published genome-wide association studies between the OR of genetic variants associated with a particular disease and the prevalence of that disease.

View Article: PubMed Central - PubMed

Affiliation: Cologne Center for Genomics, University of Cologne, Cologne, Germany; Institute of Epidemiology, Christian-Albrechts University, Kiel, Germany.

ABSTRACT
Guided by the practice of classical epidemiology, research into the genetic basis of complex disease has usually taken for granted the dictum that causative mutations are invariably over-represented among clinically affected as compared to unaffected individuals. However, we show that this supposition is not true and that a mutation contributing to the etiology of a complex disease can, under certain circumstances, be depleted among patients. Populations with defined disease prevalence were repeatedly simulated under a Wright-Fisher model, assuming various types of population history and genotype-phenotype relationship. For each simulation, the resulting mutation-specific population frequencies and odds ratios (ORs) were evaluated. In addition, the relationship between mutation frequency and OR was studied using real data from the NIH GWAS catalogue of reported phenotype associations of single-nucleotide polymorphisms (SNPs). While rare diseases (prevalence <1%) were found to be consistently caused by rare mutations with ORs>1, up to 20% of mutations causing a pandemic disease (prevalence 10-20%) had ORs<1, and their population frequency ranged from 0% to 100%. Moreover, simulation-based ORs exhibited a wide distribution, irrespective of mutation frequency. In conclusion, a substantial proportion of mutations causing common complex diseases may appear 'protective' in genetic epidemiological studies and hence would normally tend to be excluded, albeit erroneously, from further study. This apparently paradoxical result is explicable in terms of mutual confounding of the respective genotype-phenotype relationships due to a negative correlation between causal mutations induced by their common gene genealogy. As would be predicted by our findings, a significant negative correlation became apparent in published genome-wide association studies between the OR of genetic variants associated with a particular disease and the prevalence of that disease.

No MeSH data available.


Related in: MedlinePlus