Hierarchical clustering is frequently used for grouping results in expression or haplotype analyses. These methods can elucidate patterns between measures that can then be applied to discerning their validity in discriminating between experimental conditions. Here a hierarchical clustering method is used to analyze the results of an imaging genetics study using multiple brain morphology and cognitive testing endpoints for older adults with amnestic mild cognitive impairment (MCI) or cognitive complaints (CC) compared to healthy controls (HC). The single nucleotide polymorphisms (SNPs) are a subset of those included on a larger array that are found in a reported Alzheimer's disease (AD) and neurodegeneration pathway. The results indicate that genetic models within the endpoints cluster together, while there are 4 distinct sets of SNPs that differentiate between the endpoints, with most significant results associated with morphology endpoints rather than cognitive testing of patients' reported symptoms. The genes found in at least one cluster are ABCB1, APBA1, BACE1, BACE2, BCL2, BCL2L1, CASP7, CHAT, CST3, DRD3, DRD5, IL6, LRP1, NAT1, and PSEN2. The greater associations with morphology endpoints suggests that changes in brain structure can be influenced by an individual's genetic background in the absence of dementia and in some cases (Cognitive Complaints group) even without those effects necessarily being detectable on commonly used clinical tests of cognition. The results are consistent with polygenic influences on early neurodegenerative changes and demonstrate the effectiveness of hierarchical clustering in identifying genetic associations among multiple related phenotypic endpoints.
(c) 2010 Wiley-Liss, Inc.