Even so, others have argued that this might be an obvious but not in itself sufficient argument for needing very large sample sizes, as it does not take into account the increase in effect sizes enabled by careful selection of phenotypes (this is an area of neuroimaging research in itself; see Table 1), nor corroborative evidence from other sets of data, nor independent replication of top hits—which is commonly performed in GWAS studies. Clearly, an alternative way to avoid false positives is to collect additional functional evidence for a variant (the “convergent approach”). A final approach is to focus on candidate variants with known molecular function, and there is a long tradition of work relating genetic variants in the monoamine neurotransmitter pathways to risk for psychiatric illness. For instance, lower hippocampal but larger amygdala volumes have been associated with the long variant of the serotonin transporter polymorphism in major depression (Frodl et al. 2004, 2008), and variants in the 5-HT1A receptor gene have been related to amygdala volume in borderline personality disorder (Zetzsche et al. 2008). Variants in the Huntington’s disease gene have also been shown to affect normal brain structure (Mühlau et al. 2012).
