Disease risk/resilience (DR) prediction requires statistical models that are typically generated from empirical studies. For commonly occurring polygenically inherited conditions such as gout, type 2 diabetes, and cardiovascular conditions, risk/resilience estimates have most often been derived from GWAS (genome-wide association studies). Such studies require large sample sizes (n > 104 participants) genotyped with 104-107 DNA markers.
These datasets often do not include indigenous peoples, who can have important genetic differences from more commonly represented populations of predominantly European descent. Moreover, existing datasets from Māori (and Pasifika) domiciled in New Zealand are few, and those that could be utilised consist of fewer than two thousand individuals - and thus are underpowered for clinically accurate DR prediction. In addition, establishing sufficiently large GWAS is unlikely in Aotearoa/NZ because of substantive costs associated with generating genotypic data and reluctance of many Māori to participate in such studies.
In order to offset further health inequities arising from a lack of Māori-specific DR prediction models, new studies are required. Such studies require both (a) optimal designs that incorporate known genetic relationships on non-genotyped as well as genotyped individuals, and (b) analytical methods that more accurately predict phenotype than GWAS-based methods such as polygenic risk score (PRS).
We have used a population simulator (SLiM) to model genetic structures of Māori communities (i.e., whānau/hapū/iwi), incorporating estimates of effective population sizes prior to European admixture, as well as post-colonisation admixture with Europeans. We are using these simulations to explore what features of study design and analytical methods lead to optimal DR prediction. I will illustrate and present current results on this.