Machine learning approaches for crop improvement: Leveraging phenotypic and genotypic big data

Highly efficient and accurate selection of elite genotypes can lead to dramatic shortening of the breeding cycle in major crops relevant for sustaining present demands for food, feed, and fuel. In contrast to classical approaches that emphasize the need for resource-intensive phenotyping at all stages of artificial selection, genomic selection dramatically reduces the need for phenotyping. Genomic selection relies on advances in machine learning and the availability of genotyping data to predict agronomically relevant phenotypic traits. Here we provide a systematic review of machine learning approaches applied for genomic selection of single and multiple traits in major crops in the past decade. We emphasize the need to gather data on intermediate phenotypes, e.g. metabolite, protein, and gene expression levels, along with developments of modeling techniques that can lead to further improvements of genomic selection. In addition, we provide a critical view of factors that affect genomic selection, with attention to transferability of models between different environments. Finally, we highlight the future aspects of integrating high-throughput molecular phenotypic data from omics technologies with biological networks for crop improvement.

Target-oriented prioritization: targeted selection strategy by integrating organismal and molecular traits through predictive analytics in breeding

Genomic prediction in crop breeding is hindered by modeling on limited phenotypic traits. We propose an integrative multi-trait breeding strategy via machine learning algorithm, target-oriented prioritization (TOP). Using a large hybrid maize population, we demonstrate that the accuracy for identifying a candidate that is phenotypically closest to an ideotype, or target variety, achieves up to 91%. The strength of TOP is enhanced when omics level traits are included. We show that TOP enables selection of inbreds or hybrids that outperform existing commercial varieties. It improves multiple traits and accurately identifies improved candidates for new varieties, which will greatly influence breeding.

Context-dependent female preference for multiple ornaments in the bearded reedling

While it is well established that females prefer to mate with well‐ornamented males, the influence of perceptive and cognitive processes on the expression of female mate choice is still poorly known. It has been suggested that the female perception of a male's attractiveness is not absolute, but depends on the other males with which he is compared that have been previously encountered (comparative evaluation). We investigated whether mate preference in bearded reedlings (Panurus biarmicus) is dependent on or independent of social context in relation to two different traits: beard and tail lengths. Each female had a choice between two to three males with different modifications of beard and tail. For each female, three different experiments were conducted (one binary and two trinary tests). We found that when females are presented with options that vary antagonistically with respect to two ornaments (binary test), some individuals prefer one trait while others the other trait. This indicates that in our bearded reedlings population exists a mate preference polymorphisms. Moreover, we found that the presence of a third stimulus, irrespective of the initial preference, reduced the strength of the initial preference – what we can call a “preference dilution effect.” Our results suggest that the female's choice may be constrained by her cognitive abilities when she is simultaneously presented with several options varying for two uncorrelated traits.

Multiple genotype-phenotype association study reveals intronic variant pair on SIDT2 associated with metabolic syndrome in a Korean population

Background

Metabolic syndrome is a risk factor for type 2 diabetes and cardiovascular disease. We identified common genetic variants that alter the risk for metabolic syndrome in the Korean population. To isolate these variants, we conducted a multiple-genotype and multiple-phenotype genome-wide association analysis using the family-based quasi-likelihood score (MFQLS) test. For this analysis, we used 7211 and 2838 genotyped study subjects for discovery and replication, respectively. We also performed a multiple-genotype and multiple-phenotype analysis of a gene-based single-nucleotide polymorphism (SNP) set.

Results

We found an association between metabolic syndrome and an intronic SNP pair, rs7107152 and rs1242229, in SIDT2 gene at 11q23.3. Both SNPs correlate with the expression of SIDT2 and TAGLN, whose products promote insulin secretion and lipid metabolism, respectively. This SNP pair showed statistical significance at the replication stage.

Conclusions

Our findings provide insight into an underlying mechanism that contributes to metabolic syndrome.

Electronic supplementary material

The online version of this article (10.1186/s40246-018-0180-4) contains supplementary material, which is available to authorized users.

Modeling Multiple Responses via Bootstrapping Margins with an Application to Genetic Association Testing

The need for analysis of multiple responses arises from many applications. In behavioral science, for example, comorbidity is a common phenomenon where multiple disorders occur in the same person. The advantage of jointly analyzing multiple correlated responses has been examined and documented. Due to the difficulties of modeling multiple responses, nonparametric tests such as generalized Kendall's Tau have been developed to assess the association between multiple responses and risk factors. These procedures have been applied to genomewide association studies of multiple complex traits. Unfortunately, those nonparametric tests only provide the significance of the association but not the magnitude. We propose a Gaussian copula model with discrete margins for modeling multivariate binary responses. This model separates marginal effects from between-trait correlations. We use a bootstrapping margins approach to constructing Wald's statistic for the association test. Although our derivation is based on the fully parametric Gaussian copula framework for simplicity, the underlying assumptions to apply our method can be weakened. The bootstrapping margins approach only requires the correct specification of the model margins. Our simulation and real data analysis demonstrate that our proposed method not only increases power over some existing association tests, but also provides further insight into genetic association studies of multivariate traits.

Association detection between multiple traits and rare variants based on family data via a nonparametric method

Background

The rapid development of next-generation sequencing technologies allow people to analyze human complex diseases at the molecular level. It has been shown that rare variants play important roles for human diseases besides common variants. Thus, effective statistical methods need to be proposed to test for the associations between traits (e.g., diseases) and rare variants. Currently, more and more rare genetic variants are being detected throughout the human genome, which demonstrates the possibility to study rare variants. Yet complex diseases are usually measured as a variety of forms, such as binary, ordinal, quantitative, or some mixture of them. Therefore, the genetic mapping problem can be attributable to the association detection between multiple traits and multiple loci, with sufficiently considering the correlated structure among multiple traits.

Methods

In this article, we construct a new non-parametric statistic by the generalized Kendall's τ theory based on family data. The new test statistic has an asymptotic distribution, it can be used to study the associations between multiple traits and rare variants, which broadens the way to identify genetic factors of human complex diseases.

Results

We apply our method (called Nonp-FAM) to analyze simulated data and GAW17 data, and conduct comprehensive comparison with some existing methods. Experimental results show that the proposed family-based method is powerful and robust for testing associations between multiple traits and rare variants, even if the data has some population stratification effect.

Genotype × environment interactions for potato yield and quality traits: Identification of ideotypes adapted in different ecological regions of Northwest China

BACKGROUND

As a multi-use cash crop, the yield and quality traits of potatoes are often affected by genotype × environment interactions (GEI). Understanding the influence of GEI on potato yield and quality across varying environmental conditions is crucial for selecting excellent varieties suitable for specific ecological regions.

METHODS

Ten advanced potato lines and twenty-two cultivars were assessed at three pilot sites in Gansu Province, China over two successive years (2020-2021). Tuber yield, dry matter, starch, reducing sugar, carotenoid, vitamin C, pyridoxamine, thiamine, nicotinic acid, pyridoxal and pyridoxine were analyzed.

RESULTS

An analysis of variance (ANOVA) showed significant effects of GEI on all traits, with the greatest effects observed for vitamin C, reducing sugars, and carotenoids. AMMI, the additive main effects and multiplicative interaction (AMMI), and the genotype main effect plus genotype × environment interaction (GGE) biplot analysis revealed that G25 exhibited high yield, high dry matter content, and high carotenoid content in Weiyuan County. In Anding District, G18 showed high yield, high thiamine content, and low reducing sugar content. Meanwhile, G7 demonstrated superior performance with high yield, high carotenoid and thiamine contents, along with high pyridoxine content in Yongchang County. Genotype recommendations based on the mean values and stability of a single trait are partial and prejudiced, while selections based on multiple traits are desirable. Five ideal genotypes (G23, G26, G25, G24, and G16) were selected by multi-trait stability index (MTSI) in consideration of both the mean values and stability of yield and key quality traits.

CONCLUSIONS

Multi-environment trials revealed that GEI significantly affected vitamin C, reducing sugars, and carotenoids in potato. Five high-yielding and high-quality potato genotypes adapted to different ecological regions of Gansu Province were identified using MTSI analysis. This study provides references for the regional cultivation and quality breeding of potatoes.

Beyond the Standard GWAS-A Guide for Plant Biologists

Classic genome-wide association studies (GWAS) look for associations between individual single-nucleotide polymorphisms (SNPs) and phenotypes of interest. With the rapid progress of high-throughput genotyping and phenotyping technologies, GWAS have become increasingly powerful for detecting genetic determinants and their molecular mechanisms underpinning natural phenotypic variation. However, GWAS frequently yield results with neither expected nor promising loci, nor any significant associations. This is often because associations between SNPs and a single phenotype are confounded, for example with the environment, other traits or complex genetic structures. Such confounding can mask true genotype-phenotype associations, or inflate spurious associations. To address these problems, numerous methods have been developed that go beyond the standard model. Such advanced GWAS models are flexible and can offer improved statistical power for understanding the genetics underlying complex traits. Despite this advantage, these models have not been widely adopted and implemented compared to the standard GWAS approach, partly because this literature is diverse and often technical. In this review, our aim is to provide an overview of the application and the benefits of various advanced GWAS models for handling complex traits and genetic structures, targeting plant biologists who wish to carry out GWAS more effectively.

Quantitative trait locus mapping analysis of multiple traits when using genotype data with potential errors

Background

Quantitative trait locus (QTL) analysis aims to locate and estimate the effects of the genes influencing quantitative traits and infer the relationship between gene variants and changes in phenotypic characteristics using statistical methods. Some methods have been developed to map QTLs of multiple traits in the case of no genotype error in a given dataset. However, practical genetic data that people use may contain some potential errors because of the limitations of biotechnology. Common genetic data correction methods can only reduce errors, but cannot calculate the degree of error. In this paper, we propose a QTL mapping strategy for multiple traits in the presence of genotype errors.

Methods

The additive effect, dominant effect, recombination rate, error rate, and other parameters of QTLs can be simultaneously obtained using this new method in the framework of multiple-interval mapping.

Results

Our simulation results show that the accuracy of parameter estimation can be improved by considering the errors of marker genotypes during the analysis of genetic data. Real data analysis also shows that the new method proposed in this paper can map the QTLs of multiple traits more accurately.