Selecting subsets of genotyped experimental populations for phenotyping to maximize genetic diversity

Selective phenotyping is a way of capturing the benefits of large population sizes without the need to carry out large-scale phenotyping and hence is a cost-effective means of capturing information about gene-trait relationships within a population. The diversity within the sample gives an indication of the efficiency of this information capture; less diversity implies greater redundancy of the genetic information. Here, we propose a method to maximize genetic diversity within the selected samples. Our method is applicable to general experimental designs and robust to common problems such as missing data and dominant markers. In particular, we discuss its application to multi-parent advanced generation intercross (MAGIC) populations, where, although thousands of lines may be genotyped as a large population resource, only hundreds may need to be phenotyped for individual studies. Through simulation, we compare our method to simple random sampling and the minimum moment aberration method. While the gain in power over simple random sampling for all tested methods is not large, our method results in a much more diverse sample of genotypes. This diversity can be applied to improve fine mapping resolution once a QTL region has been detected. Further, when applied to two wheat datasets from doubled haploid and MAGIC progeny, our method detects known QTL for small sample sizes where other methods fail.