Genotyping Platforms for Genome-Wide Association Studies: Options and Practical Considerations.

Genome-wide association studies (GWAS) in crops requires genotyping platforms that are capable of producing accurate high density genotyping data on hundreds of plants in a cost-effective manner. Currently there are multiple commercial platforms available that are being effectively used across crops. These platforms include genotyping arrays such as the Illumina Infinium arrays and the Applied Biosystems Axiom Arrays along with a variety of resequencing methods. These methods are being used to genotype tens of thousands of markers up to millions of markers on GWAS panels. They are being used on crops with simple genomes to crops with very complex, large, polyploid genomes. Depending on the crop and the goal of the GWAS, there are several options and practical considerations to take into account when selecting a genotyping technology to ensure that the right coverage, accuracy, and cost for the study is achieved.

Multi-Population Selective Genotyping to Identify Soybean [Glycine max (L.) Merr.] Seed Protein and Oil QTLs.

Plant breeders continually generate ever-higher yielding cultivars, but also want to improve seed constituent value, which is mainly protein and oil, in soybean [Glycine max (L.) Merr.]. Identification of genetic loci governing those two traits would facilitate that effort. Though genome-wide association offers one such approach, selective genotyping of multiple biparental populations offers a complementary alternative, and was evaluated here, using 48 F2:3 populations (n = âˆ¼224 plants) created by mating 48 high protein germplasm accessions to cultivars of similar maturity, but with normal seed protein content. All F2:3 progeny were phenotyped for seed protein and oil, but only 22 high and 22 low extreme progeny in each F2:3 phenotypic distribution were genotyped with a 1536-SNP chip (ca 450 bimorphic SNPs detected per mating). A significant quantitative trait locus (QTL) on one or more chromosomes was detected for protein in 35 (73%), and for oil in 25 (52%), of the 48 matings, and these QTL exhibited additive effects of ≥ 4 g kg(-1) and R(2) values of 0.07 or more. These results demonstrated that a multiple-population selective genotyping strategy, when focused on matings between parental phenotype extremes, can be used successfully to identify germplasm accessions possessing large-effect QTL alleles. Such accessions would be of interest to breeders to serve as parental donors of those alleles in cultivar development programs, though 17 of the 48 accessions were not unique in terms of SNP genotype, indicating that diversity among high protein accessions in the germplasm collection is less than what might ordinarily be assumed.

A reference genome for common bean and genome-wide analysis of dual domestications.

Common bean (Phaseolus vulgaris L.) is the most important grain legume for human consumption and has a role in sustainable agriculture owing to its ability to fix atmospheric nitrogen. We assembled 473 Mb of the 587-Mb genome and genetically anchored 98% of this sequence in 11 chromosome-scale pseudomolecules. We compared the genome for the common bean against the soybean genome to find changes in soybean resulting from polyploidy. Using resequencing of 60 wild individuals and 100 landraces from the genetically differentiated Mesoamerican and Andean gene pools, we confirmed 2 independent domestications from genetic pools that diverged before human colonization. Less than 10% of the 74 Mb of sequence putatively involved in domestication was shared by the two domestication events. We identified a set of genes linked with increased leaf and seed size and combined these results with quantitative trait locus data from Mesoamerican cultivars. Genes affected by domestication may be useful for genomics-enabled crop improvement.

Identification and validation of quantitative trait loci for seed yield, oil and protein contents in two recombinant inbred line populations of soybean.

Soybean seeds contain high levels of oil and protein, and are the important sources of vegetable oil and plant protein for human consumption and livestock feed. Increased seed yield, oil and protein contents are the main objectives of soybean breeding. The objectives of this study were to identify and validate quantitative trait loci (QTLs) associated with seed yield, oil and protein contents in two recombinant inbred line populations, and to evaluate the consistency of QTLs across different environments, studies and genetic backgrounds. Both the mapping population (SD02-4-59 × A02-381100) and validation population (SD02-911 × SD00-1501) were phenotyped for the three traits in multiple environments. Genetic analysis indicated that oil and protein contents showed high heritabilities while yield exhibited a lower heritability in both populations. Based on a linkage map constructed previously with the mapping population and using composite interval mapping and/or interval mapping analysis, 12 QTLs for seed yield, 16 QTLs for oil content and 11 QTLs for protein content were consistently detected in multiple environments and/or the average data over all environments. Of the QTLs detected in the mapping population, five QTLs for seed yield, eight QTLs for oil content and five QTLs for protein content were confirmed in the validation population by single marker analysis in at least one environment and the average data and by ANOVA over all environments. Eight of these validated QTLs were newly identified. Compared with the other studies, seven QTLs for seed yield, eight QTLs for oil content and nine QTLs for protein content further verified the previously reported QTLs. These QTLs will be useful for breeding higher yield and better quality cultivars, and help effectively and efficiently improve yield potential and nutritional quality in soybean.

A genome-wide association study of seed protein and oil content in soybean.

BACKGROUND: Association analysis is an alternative to conventional family-based methods to detect the location of gene(s) or quantitative trait loci (QTL) and provides relatively high resolution in terms of defining the genome position of a gene or QTL. Seed protein and oil concentration are quantitative traits which are determined by the interaction among many genes with small to moderate genetic effects and their interaction with the environment. In this study, a genome-wide association study (GWAS) was performed to identify quantitative trait loci (QTL) controlling seed protein and oil concentration in 298 soybean germplasm accessions exhibiting a wide range of seed protein and oil content. RESULTS: A total of 55,159 single nucleotide polymorphisms (SNPs) were genotyped using various methods including Illumina Infinium and GoldenGate assays and 31,954 markers with minor allele frequency >0.10 were used to estimate linkage disequilibrium (LD) in heterochromatic and euchromatic regions. In euchromatic regions, the mean LD (r2) rapidly declined to 0.2 within 360 Kbp, whereas the mean LD declined to 0.2 at 9,600 Kbp in heterochromatic regions. The GWAS results identified 40 SNPs in 17 different genomic regions significantly associated with seed protein. Of these, the five SNPs with the highest associations and seven adjacent SNPs were located in the 27.6-30.0 Mbp region of Gm20. A major seed protein QTL has been previously mapped to the same location and potential candidate genes have recently been identified in this region. The GWAS results also detected 25 SNPs in 13 different genomic regions associated with seed oil. Of these markers, seven SNPs had a significant association with both protein and oil. CONCLUSIONS: This research indicated that GWAS not only identified most of the previously reported QTL controlling seed protein and oil, but also resulted in narrower genomic regions than the regions reported as containing these QTL. The narrower GWAS-defined genome regions will allow more precise marker-assisted allele selection and will expedite positional cloning of the causal gene(s).

Structural and functional divergence of a 1-Mb duplicated region in the soybean (Glycine max) genome and comparison to an orthologous region from Phaseolus vulgaris.

Soybean (Glycine max) has undergone at least two rounds of polyploidization, resulting in a paleopolyploid genome that is a mosaic of homoeologous regions. To determine the structural and functional impact of these duplications, we sequenced two ~1-Mb homoeologous regions of soybean, Gm8 and Gm15, derived from the most recent ~13 million year duplication event and the orthologous region from common bean (Phaseolus vulgaris), Pv5. We observed inversions leading to major structural variation and a bias between the two chromosome segments as Gm15 experienced more gene movement (gene retention rate of 81% in Gm15 versus 91% in Gm8) and a nearly twofold increase in the deletion of long terminal repeat (LTR) retrotransposons via solo LTR formation. Functional analyses of Gm15 and Gm8 revealed decreases in gene expression and synonymous substitution rates for Gm15, for instance, a 38% increase in transcript levels from Gm8 relative to Gm15. Transcriptional divergence of homoeologs was found based on expression patterns among seven tissues and developmental stages. Our results indicate asymmetric evolution between homoeologous regions of soybean as evidenced by structural changes and expression variances of homoeologous genes.

Genome sequence of the palaeopolyploid soybean.

Soybean (Glycine max) is one of the most important crop plants for seed protein and oil content, and for its capacity to fix atmospheric nitrogen through symbioses with soil-borne microorganisms. We sequenced the 1.1-gigabase genome by a whole-genome shotgun approach and integrated it with physical and high-density genetic maps to create a chromosome-scale draft sequence assembly. We predict 46,430 protein-coding genes, 70% more than Arabidopsis and similar to the poplar genome which, like soybean, is an ancient polyploid (palaeopolyploid). About 78% of the predicted genes occur in chromosome ends, which comprise less than one-half of the genome but account for nearly all of the genetic recombination. Genome duplications occurred at approximately 59 and 13 million years ago, resulting in a highly duplicated genome with nearly 75% of the genes present in multiple copies. The two duplication events were followed by gene diversification and loss, and numerous chromosome rearrangements. An accurate soybean genome sequence will facilitate the identification of the genetic basis of many soybean traits, and accelerate the creation of improved soybean varieties.

Application of machine learning in SNP discovery.

BACKGROUND: Single nucleotide polymorphisms (SNP) constitute more than 90% of the genetic variation, and hence can account for most trait differences among individuals in a given species. Polymorphism detection software PolyBayes and PolyPhred give high false positive SNP predictions even with stringent parameter values. We developed a machine learning (ML) method to augment PolyBayes to improve its prediction accuracy. ML methods have also been successfully applied to other bioinformatics problems in predicting genes, promoters, transcription factor binding sites and protein structures. RESULTS: The ML program C4.5 was applied to a set of features in order to build a SNP classifier from training data based on human expert decisions (True/False). The training data were 27,275 candidate SNP generated by sequencing 1973 STS (sequence tag sites) (12 Mb) in both directions from 6 diverse homozygous soybean cultivars and PolyBayes analysis. Test data of 18,390 candidate SNP were generated similarly from 1359 additional STS (8 Mb). SNP from both sets were classified by experts. After training the ML classifier, it agreed with the experts on 97.3% of test data compared with 7.8% agreement between PolyBayes and experts. The PolyBayes positive predictive values (PPV) (i.e., fraction of candidate SNP being real) were 7.8% for all predictions and 16.7% for those with 100% posterior probability of being real. Using ML improved the PPV to 84.8%, a 5- to 10-fold increase. While both ML and PolyBayes produced a similar number of true positives, the ML program generated only 249 false positives as compared to 16,955 for PolyBayes. The complexity of the soybean genome may have contributed to high false SNP predictions by PolyBayes and hence results may differ for other genomes. CONCLUSION: A machine learning (ML) method was developed as a supplementary feature to the polymorphism detection software for improving prediction accuracies. The results from this study indicate that a trained ML classifier can significantly reduce human intervention and in this case achieved a 5-10 fold enhanced productivity. The optimized feature set and ML framework can also be applied to all polymorphism discovery software. ML support software is written in Perl and can be easily integrated into an existing SNP discovery pipeline.