Sequencing depth and genotype quality: Accuracy and breeding operation considerations for genomic selection applications in autopolyploid crops

Abstract

The autopolyploid nature of potato and sweetpotato ensures a wide range of meiotic configurations and linkage phases leading to complex gene action and pose problems in genotype data quality and genomic selection analyses. We used a 315-progeny biparental population of hexaploid sweetpotato and a diversity panel of 380 tetraploid potato, genotyped using different platforms to answer the following questions: i) do polyploid crop breeders need to invest more for additional sequencing depth? ii) how many markers are required to make selection decisions? iii) does considering non-additive genetic effects improve predictive ability (PA)? iv) does considering dosage or quantitative trait loci (QTL) offer significant improvement to PA? Our results show that only a small number of highly informative single nucleotide polymorphisms (SNPs; ≤ 1000) are adequate for prediction, hence it is possible to get this number at the current sequencing depth from most service providers. We also show that considering dosage information and additive-effects only models had the best PA for most traits, while the comparative advantage of considering non-additive genetic effects and including known QTL in the predictive model depended on trait architecture. We conclude that genomic selection can help accelerate the rate of genetic gains in potato and sweetpotato. However, application of genomic selection should be considered as part of optimizing the entire breeding program. Additionally, since the predictions in the current study are based on single populations, further studies on the effects of haplotype structure and inheritance on PA should be studied in actual multi-generation breeding populations.Key messagePolypoid crop breeders do not need more investment for sequencing depth, dosage information and fewer highly informative SNPs recommended, non-additive models and QTL advantages on prediction dependent on trait architecture.

Development of diagnostic SNP markers for quality assurance and control in sweetpotato [Ipomoea batatas (L.) Lam.] breeding programs

“…We chose to use 'diploidized' data in the current analysis because the depth of coverage from most genotyping platforms is not adequate to reliably characterize heterozygous loci, such as those likely to be found in polyploids, for which deeper sequencing is required [43]. Furthermore, analyses comparing genotypic data from DArTSeq and those from a deep sequencing optimization platform for sweetpotato called GBSpoly [44] have confirmed that highly informative 'diploidized' DArTseq data performed just as well as high confidence data with dosage in genomic predictions of sweetpotato for traits with simple trait architecture [20]. This therefore implies…”

“…Sequencing was done at 96-plex, high density and SNP calling done using DArT's proprietary software DArTSoft [18], with aligning to the diploid reference genome of Ipomoea trifida, a relative of sweetpotato [19]. Given that most commercial genotyping platforms have allele depth coverage~25x to 30x, previous studies [20] have shown that this depth of coverage is not adequate to call allele dosage with confidence in genotype quality for hexaploid sweetpotato. The study also showed that in such cases, 'diploidized' biallelic loci which are informative enough performed almost as well as data with high confidence dosage information, for simple traits.…”

Quantitative Trait Locus Mapping for Common Scab Resistance in a Tetraploid Potato Full-Sib Population

“…(Enciso- Rodriguez et al, 2018). Haplotypic and QTL information, such as that provided here, can be used to leverage genomic-assisted prediction models in order to deliver higher predictive abilities, notably for less complex traits (Gemenet et al, 2020). (Takken et al, 2006), and transcription factors such as NAC (PGSC0003DMT400097372, ~54.15 Mbp; PGSC0003DMT400096000, ~54.17 Mbp) (Nuruzzaman et al, 2013) and WRKY (PGSC0003DMT400046570, ~53.04) (Bhattarai et al, 2010;Enciso-Rodriguez et al, 2018).…”

Quantitative trait locus mapping for common scab resistance in a tetraploid potato full-sib population

“…For potato common scab, such genomic-assisted prediction models have shown prediction accuracies as high as 0.278, and a SNP with major effect on chromosome 9 (Enciso-Rodriguez et al, 2018). Haplotypic and QTL information, such as that provided here, can be used to leverage genomic-assisted prediction models in order to deliver higher predictive abilities, notably for less complex traits (Gemenet et al, 2020).…”