Schizophrenia: the testing of genetic models by pedigree analysis.

Simulated pedigrees of schizophrenia generally show a clear peak in their likelihood surface corresponding to analysis by the genetic models, which served as the basis for the simulation. The likelihood surface obtained with real data permits determination of the allelic frequency and the selection of an optimal one-locus, two-locus, and four-locus model. These three models have certain features in common, notably, a relatively high frequency of the allele predisposing to schizophrenia (about 20%) and a relatively low index of genetic determination (23%--34%). However, direct likelihood comparisons do not permit distinctions between the one-locus, two-locus, and four-locus models. The most likely interpretation of this finding is that the etiology of schizophrenia is heterogeneous or even nongenetic. However, a simple model with a single completely recessive locus and incomplete penetrance in the homozygote also produces a flat likelihood surface closely resembling that obtained with the real data. With reservation, this single-locus model may be put forward as a potentially useful working hypothesis.