The Addition of Epistatic Genetic Effects Increases Genomic Prediction Accuracy for Reproduction and Production Traits in Duroc Pigs Using Genomic Models

Background: Genomic selection has been implemented in livestock genetic evaluations for years. However, currently most genomic selection models only consider the additive effects associated with SNP markers and nonadditive genetic effects have been for the most part ignored. Methods: Production traits for 26,735 to 27,647 Duroc pigs and reproductive traits for 5,338 sows were used, including off-test body weight (WT), off-test back fat (BF), off-test loin muscle depth (MS), number born alive (NBA), number born dead (NBD), and number weaned (NW). All animals were genotyped with the PorcineSNP60K Bead Chip. Variance components were estimated using a linear mixed model that includes inbreeding coefficient, additive, dominance, additive-by-additive, additive-by-dominance, dominance-by-dominance effect, and common litter environmental effect. Genomic prediction performance, including all nonadditive genetic effects, was compared with a reduced model that included only additive genetic effect. Results: Significant estimates of additive-by-additive effect variance were observed for NBA, BF, and WT (31%, 9%, and 10%, respectively). Production traits showed significant large estimates of additive-by-dominance variance (9%-23%). MS also showed large estimate of dominance-by-dominance variance (10%). Dominance effect variance estimates were low for all traits (0%-2%). Compared to the reduced model, prediction accuracies using the full model, including nonadditive effects, increased significantly by 12%, 12%, and 1% for NBA, WT, and MS, respectively. A strong dominance association signal with BF was identified near AK5.Conclusions: Sizable estimates of epistatic effects were found for the reproduction and production traits, while the dominance effect was relatively small for all traits yet significant for all production traits. Including nonadditive effects, especially epistatic effects in the genomic prediction model, significantly improved prediction accuracy for NBA, WT, and MS.

The impact of epistasis in the heterosis and combining ability analyses

The current theoretical knowledge concerning the influence of epistasis on heterosis is based on simplified multiplicative model. The objective of this study was to assess the impact of epistasis in the heterosis and combining ability analyses, assuming additive model, hundreds of genes, linkage disequilibrium (LD), dominance, and seven types of digenic epistasis. We developed the quantitative genetics theory for supporting the simulation of the individual genotypic values in nine populations, the selfed populations, the 36 interpopulation crosses, 180 doubled haploids (DHs) and their 16,110 crosses, assuming 400 genes in 10 chromosomes of 200 cM. Epistasis only affects population heterosis if there is LD. Only additive x additive and dominance x dominance epistasis can affect the components of the heterosis and combining ability analyses of populations. Both analyses can lead to completely wrong inferences regarding the identification of the superior populations, the populations with greater differences of gene frequencies, and the populations with maximum variability, when the number of interacting genes and the magnitude of the epistatic effects are high. There was a decrease in the average heterosis by increasing the number of epistatic genes and the magnitude of their epistatic effects. The same results are generally true for the combining ability analysis of DHs. Surprisingly, the combining ability analyses of subsets of 20 DHs showed no significant average impact of epistasis on the identification of the most divergent ones, even assuming a high number of epistatic genes and great magnitude of their effects. However, a significant negative effect can occur.

Efectele genetice implicate în răspunsul grăului comun la filtratul de cultură Drechslera sorokiniana (SACC.) subram

Helminthosporiosis caused by the fungus Drechslera sorokiniana (Sacc.) causes significant crop and quality losses to Triticum aestivum L. in agroecological conditions with extreme humidity. Increasing the resistance is considered the most cost-effective and sustainable approach to disease control. The aim of this study was to determine the genetic effects involved in the inheritance of resistance, using the ge-netic model of character reproduction in descendants of wheat. Generations F1, F2, BCP1 and BCP2, de-scended from the mutual crossing of the parents Basarabeanca / Moldova 30 and Moldova 30 / Moldova 3 (P1 and P2) were evaluated for the response of callus characters to the action of D. sorokiniana culture filtrate on the medium Murashige Skoog. Fungal metabolites have decreased the effects of gene actions and epistatic interactions, but also their variance. The phenomenon corresponds to the decrease of callus indices. A great importance for the heredity of the character of the surface of the callus manifested the epistatic effects of additive-dominant (ad) type. In the case of callus biomass comparable to the mean val-ues were the a actions, but also the ad and dd epistatic effects. The predominant involvement of epistatic effects indicates the need for resistance selections to helminthosporiosis in late generations of wheat.

Understanding the Effectiveness of Genomic Prediction in Tetraploid Potato

Use of genomic prediction (GP) in tetraploid is becoming more common. Therefore, we think it is the right time for a comparison of GP models for tetraploid potato. GP models were compared that contrasted shrinkage with variable selection, parametric vs. non-parametric models and different ways of accounting for non-additive genetic effects. As a complement to GP, association studies were carried out in an attempt to understand the differences in prediction accuracy. We compared our GP models on a data set consisting of 147 cultivars, representing worldwide diversity, with over 39 k GBS markers and measurements on four tuber traits collected in six trials at three locations during 2 years. GP accuracies ranged from 0.32 for tuber count to 0.77 for dry matter content. For all traits, differences between GP models that utilised shrinkage penalties and those that performed variable selection were negligible. This was surprising for dry matter, as only a few additive markers explained over 50% of phenotypic variation. Accuracy for tuber count increased from 0.35 to 0.41, when dominance was included in the model. This result is supported by Genome Wide Association Study (GWAS) that found additive and dominance effects accounted for 37% of phenotypic variation, while significant additive effects alone accounted for 14%. For tuber weight, the Reproducing Kernel Hilbert Space (RKHS) model gave a larger improvement in prediction accuracy than explicitly modelling epistatic effects. This is an indication that capturing the between locus epistatic effects of tuber weight can be done more effectively using the semi-parametric RKHS model. Our results show good opportunities for GP in 4x potato.

Prediction performance of linear models and gradient boosting machine on complex phenotypes in outbred mice

Recent literature suggests machine learning methods can capture interactions between loci and therefore could outperform linear models when predicting traits with relevant epistatic effects. However, investigating this empirically requires data with high mapping resolution and phenotypes for traits with known non-additive gene action. The objective of the present study was to compare the performance of linear (GBLUP, BayesB and elastic net [ENET]) methods to a non-parametric tree-based ensemble (gradient boosting machine GBM) method for genomic prediction of complex traits in mice. The dataset used contained phenotypic and genotypic information for 835 animals from 6 non-overlapping generations. Traits analyzed were bone mineral density (BMD), body weight at 10, 15 and 20 weeks (BW10, BW15 and BW20), fat percentage (FAT%), circulating cholesterol (CHOL), glucose (GLUC), insulin (INS) and triglycerides (TGL), and urine creatinine (UCRT). After quality control, the genotype dataset contained 50,112 SNP markers. Animals from older generations were considered as a reference subset, while animals in the latest generation as candidates for the validation subset. We also evaluated the impact of different levels of connectedness between reference and validation sets. Model performance was measured as the Pearsons correlation coefficient and mean squared error (MSE) between adjusted phenotypes and the models prediction for animals in the validation subset. Outcomes were also compared across models by checking the overlapping top markers and animals. Linear models outperformed GBM for seven out of ten traits. For these models, accuracy was proportional to the traits heritability. For traits BMD, CHOL and GLU, the GBM model showed better prediction accuracy and lower MSE. Interestingly, for these three traits there is evidence in literature of a relevant portion of phenotypic variance being explained by epistatic effects. We noticed that for lower connectedness, i.e., imposing a gap of one to two generations between reference and validation populations, the superior performance of GBM was only maintained for GLU. Using a subset of top markers selected from a GBM model helped for some of the traits to improve accuracy of prediction when these were fitted into linear and GBM models. The GBM model showed consistently fewer markers and animals in common among the top ranked than linear models. Our results indicate that GBM is more strongly affected by data size and decreased connectedness between reference and validation sets than the linear models. Nevertheless, our results indicate that GBM is a competitive method to predict complex traits in an outbred mice population, especially for traits with assumed epistatic effects.

Identification of QTLs Associated with Two-seed Pod Length and Width in Soybean

Background: Two-seed pod length and width (TSPL and TSPW, respectively) are the traits underlying seed size, which is an important factor influencing soybean yield. Methods: A population comprising 213 chromosome segment substitution lines from a cross between ‘Suinong14’ (SN14) and ZYD00006 was used for a quantitative trait locus (QTL) analysis. The QTLs were identified on the basis of the phenotypes from 2016 to 2019. Additionally, IciMapping 4.2 was used to analyze the phenotypic and genetic data. Genes were annotated using the KEGG and Phytozome databases. Result: Five QTLs for TSPL and four QTLs for TSPW were identified. One QTL on chromosome 17 was detected for TSPL in 2017 and 2018 as well for TSPW in 2018 and 2019. Analyses of the additive × additive epistatic effects of QTLs revealed six stable loci pairs for epistatic effects on the two traits. On the basis of an alignment of the parental gene sequences and the gene annotation information, Glyma.04G188800, Glyma.11G164700, Glyma.13G132700, Glyma.17G156100 and Glyma.13G133200 were selected as candidate genes for TSPL, whereas Glyma.13G174400, Glyma.13G174700, Glyma.16G012500, Glyma.17G156100, Glyma.19G161700 and Glyma.19G161800 were selected as candidate genes for TSPW. These results may be relevant for future attempts to modify soybean seed traits.

Deep Neural Networks for Genomic Prediction Do Not Estimate Marker Effects

Genomic prediction is a promising technology for advancing both plant and animal breeding, with many different prediction models evaluated in the literature. It has been suggested that the ability of powerful nonlinear models such as deep neural networks to capture complex epistatic effects between markers offers advantages for genomic prediction. However, these methods tend not to outperform classical linear methods, leaving it an open question why this capacity to model nonlinear effects does not seem to result in better predictive capability. In this work, we propose the theory that, due to a principle called shortcut learning, deep neural networks tend to base their predictions on overall genetic relatedness, rather than on the effects of particular markers, such as epistatic effects. Using several datasets of crop plants (lentil, wheat, and Brassica carinata), we demonstrate the network's indifference to the values of the markers by showing that the same network, provided with only the locations of matches between markers for two individuals, is able to perform prediction to the same level of accuracy.