Powdery mildew, caused by Blumeria graminis f. sp. hordei (Bgh), is a worldwide disease problem on barley (Hordeum vulgare) with potentially severe impact on yield. Historically, resistance genes have been identified chiefly from cultivated lines and landraces; however, wild barley (H. vulgare subsp. spontaneum) accessions have proven to be extraordinarily rich sources of powdery mildew resistance. This study describes the characterization of a collection of 316 wild barley accessions, known as the Wild Barley Diversity Collection (WBDC), for resistance to powdery mildew and the genetic location of powdery mildew resistance loci. The WBDC was phenotyped for reaction to 40 different Bgh isolates at the seedling stage and genotyped with 10 508 molecular markers. Accessions resistant to all 40 isolates of Bgh were not found; however, three accessions (WBDC 053, 085 and 089) exhibited resistance to 38 of the isolates. Gene postulation analyses revealed that many accessions, while resistant, contained none of the 12 genes present in the Pallas near‐isogenic lines Mla1, Mla3, Mla6, Mla7, Mla9, Mla12, Mla13, Mlk1, MlLa, Mlg, Mlat and Ml(Ru2), suggesting that the accessions carry novel genes or gene combinations. A genome‐wide association study of powdery mildew resistance in the WBDC identified 21 significant marker‐trait associations that resolved into 15 quantitative trait loci. Seven of these loci have not been previously associated with powdery mildew resistance. Taken together, these results demonstrate that the WBDC is a rich source of powdery mildew resistance, and provide genetic tools for incorporating the resistance into barley breeding programmes.

When genomewide predictions are available, maize (Zea mays L.) breeders may consider foregoing first‐year phenotyping of testcrosses or, at the very least, reducing the number of locations used in phenotyping. Our objectives were to determine the equivalency between genomewide predictions and the number of locations used in phenotyping, and the extent to which genomewide predictions can reduce subsequent phenotyping in maize. For each of 21 test populations, we constructed half‐sib training populations from prior biparental populations evaluated in multiple environments. Marker data were available for 2911 single nucleotide polymorphism markers. We estimated the number of locations (LEq) for which the response to phenotypic selection was equal to the response to genomewide selection. The median analytical estimate of LEq (cross‐validation estimate of LEq in parentheses) was 1.1 (1) for yield, 1.8 (2) for moisture, and 3.0 (3) for test weight. The estimates of LEq varied widely among the test populations. We estimated the response to selection for an index that combined genomewide predictions and phenotypic data from different numbers of predictor locations (LP). The improvement in the response when LP increased from 0 (genomewide selection) to 1 was greater than the improvement in the response when LP increased from 1 to 2. This result suggested that phenotyping even at a single location captured signals that genomewide prediction did not capture. The analysis herein is helpful in designing breeding schemes that achieve a balance between the amount of genetic gain and the time and cost required to achieve such gain.