Genotyping by Sequencing on a Worldwide Collection of Maize Inbreds, Landraces and Teosintes

A large collection of maize inbreds, landraces and teosintes, approximately 45,000 samples in total, were sequenced using the genotyping by sequencing (GBS) protocol.  We aim to discover and genotype 1.5 million SNP markers for all of the samples. In order to handle the large data set, our GBS bioinformatic pipeline was redesigned using a new strategy and the HDF5 format was applied. To improve the genotyping quality, especially the alignment which is the key for SNP calling, we developed genetic mapping approaches combined with machine learning algorithms to refine sequence alignment. From the pilot experiment evaluating the new strategy, we found: 1. Alignments are quite different between aligners; 2. Aligners differ in sensitivity and accuracy, but they are often complementary; 3. Each alignment should be viewed as a hypothesis; 4. Genetic mapping is needed to verify alignments. All together, this new strategy should significantly improve SNP discovery in complex genomes.