Linkage analysis of complex disorders with multiple phenotypic categories: simulation studies and application to bipolar disorder data.

The problem of linkage analysis of disorders with multiple possible phenotypes (diagnostic spectrum) is considered. A modification is proposed to Ott's [1994] method of down-weighting the contribution of broader diagnoses by reducing penetrance ratios for affected cases. A "robust weighting" strategy considers only the robustness of a set of ratios across a range of true genetic models. Practical models for lod-score analysis will typically employ a high penetrance ratio (> 10) for "core" cases, and ratios between 2 and 5 for broader cases. Results suggest that an additive parametric analysis correlates highly with dominant, recessive and nonparametric linkage (NPL) analyses. A weighted, additive model is then applied to a modified NIMH bipolar chromosome 18 data set (Genetic Analysis Workshop 10) and compared with NPL analyses under narrow and broad diagnostic models. The weighted model performed well. The introduction of similar weights into nonparametric analyses may prove more useful.