Cindy Boer

268 | Chapter 6 ing heritability” is thought to be partly explained by rare SNVs (minor allele frequency <0.01) as the vast majority of human genetic variation is rare. In addition, other types of genetic variations besides SNVs exist, such as structural variation, which may ex- plain a portion of the “missing heritability”. As GWAS predominantly evaluates common SNVs (minor allele frequency >0.01), a proportion of the heritability will remain undis- covered. In addition, rare genetic variation is population specific, and the majority of (~80%) GWASs were performed in individuals of European descent[50]. Differences in disease incidence, prevalence and severity between populations can in part be attribut- ed to underlying genetic variation and environmental interaction[51]. It is therefore not surprising that the results of European ancestry based GWAS are not directly transfera- ble to other populations, as the underlying causal variants and effect sizes differ across populations[51]. Thus more globally inclusive data containing more genetic variation are needed, for example by whole genome sequencing globally inclusive populations. Although whole genome-sequencing data is the golden standard for identify- ing all possible genetic variations, it is considerably more expensive and computation- al intensive than genotyping arrays. The newest generation arrays, such as the Global Screening Array(GSA) or the PMDA (Thermofisher), are not only cheap, but they also increasingly cover more of the genome for different ethnic populations, including clin- ically valuable rare variants[52]. Moreover, new imputation panels (TOPMED) and im- putation methods are improving the calling of rare and population specific variants[53]. However, there still is a lack of true population specific reference panels i.e., for the Rot- terdam Study this means a “Dutch ancestry” whole genome sequencing reference pan- el, rather than a whole genome sequencing reference panel of “European Ancestry US Citizens”. Considering that many existing cohorts already have genotyping data avail- able and could improve quality simply by re-imputing data, genotyping arrays remain highly cost effective for GWASs. Regardless of the method, for successful GWAS very large sample sizes are needed to study rare genetic variation. The costs saved by using array-based genotyping instead of whole-genome-sequencing could thus be better used to increase the sample size. However, is larger always better? Large scale GWASs have undoubtedly been suc- cessful and sample size is usually the key limiting factor in GWAS discovery. Howev- er, small scale GWAS using “smart phenotyping” such as stratified or endophenotype strategies have also been very successful, and have as extra benefit that they reduce heterogeneity and improve GWAS interpretation. Although, such smaller scale studies would be even more successful with larger sample sizes. The identification of a single robust GWAS finding could already lead to significant therapeutic impact. Thus also in GWAS: quality (robust and well validated association) rules over quantity (hundreds of