Clustering methods applied to allele sharing data

Here we focus on using clustering methods to disentangle the interacting fa ctors that lead to the presentation of complex diseases. Relative pairs are placed in discrete subgroups, or classes, based upon their pattern of alle le sharing at a sequence of markers and on concomitant risk factors. The re lationship between the locus information and the affectation status of the relative pairs within each subgroup then can be assessed. Cluster analysis (CLA) and latent class analysis (LCA) were applied to sibling allele sharin g data from GAW11 simulated data, and to an existing Alzheimer's disease (A D) dataset. Both methods were able to identify markers linked to all 3 dise ase loci in the GAWI1 data. LCA and CLA also replicated regions of chromoso mes identified in an analysis of the AD data using affected-sib-pair method s. These analyses indicate that classification tools may be useful for dete cting susceptibility genes for complex traits. Genet. Epidemiol. 19(Suppl 1 ):557-563, 2000. (C) 2000 Wiley-Liss, Inc.