Genetic heterogeneity – 'the existence of two or more genetically distinct entities with essentially one and the same phenotype'24 – is a common attribute of human disease, characterizing not only the genetically complex disorders, but inherently present in monogenic mendelian conditions as well. In the latter, locus heterogeneity is usually estimated as the proportion of families linked to a given locus using a Bayesian model25 or likelihood-ratio tests.26 The approaches to resolving such heterogeneity include segregation analysis and search for subclinical markers that may index aetiological subtypes. Retinitis pigmentosa is an apt example, where the identification of mutations in over 30 genes has resulted in a genetic classification of the retinal degeneration syndromes.27 Extensive heterogeneity is the rule in many common complex disorders, including inflammatory bowel disease,28 rheumatoid arthritis,29 and osteoporosis.30 The obverse phenomenon of pleiotropy, that is, multiple phenotypes arising from one genetic factor, is equally common, for example, in demyelinating peripheral neuropathy or axonal neuropathy with vocal cord paresis caused by mutations in the same GDAP1 gene.31, 32 A range of dissimilar syndromes may be varying expressions of a single genetic defect, such as severe neonatal intestinal obstruction, bronchiectatic lung disease, idiopathic pancreatitis, and male infertility, caused by mutations in the CFTR gene.33, 34, 35 Both heterogeneity and pleiotropy are implicated in the phenotypic variation of common brain disorders,36 where their effects may be aggravated by unknown or poorly understood environmental contributions to the phenotype.37
Schizophrenia cannot be an exception from these laws. Although heterogeneity is generally acknowledged and genetic linkage analysis often performed under the assumption of heterogeneity, this is usually performed post hoc, that is, after the data have been collected, or by default, when difficult to interpret results have been obtained.38 Apart from locus and allelic heterogeneity, commonly suspected sources of 'nuisance' variance in schizophrenia include a poorly understood, poly- or oligogenic transmission; incomplete penetrance; variable phenotype expression; unknown environmental contribution; phenocopies; misspecification of the genetic model; and measurement or classification error.39, 40 Less often acknowledged, but potentially critical sources are a fallible phenotype;41 existence of latent disease subtypes that may be aetiologically different;42 and population admixture (of subtypes and ethnic variation) that could seriously compromise the power of the available analytic methods.43, 44
The likely existence of aetiologically different subtypes of the disorder (Bleuler's notion of a 'group of schizophrenias'45), is rarely considered in genetic studies, which tend to be predicated on the broad clinical diagnosis as the phenotype, implicitly assuming a unitary view of the disorder. Phenotype refinement through disaggregation into clinical subtypes, or extension by covariate quantitative traits, has been a successful strategy in the genetic dissection of asthma,20 type I diabetes,19 or dementia.23 This approach has had limited following in schizophrenia research. The failure to address the unresolved heterogeneity in schizophrenia continues to be a serious obstacle to the effective harnessing of novel technologies, such as whole-genome association analysis, or joint association and expression studies, into an effort to 'deconstruct' schizophrenia.