Search bioRxiv⌕ Search

Biology subjects

Iwata, H.

Publications and source records attributed to Iwata, H..

At least 19 recordsLinked to original sources
bioRxiv2026

Metabolome genome-wide association study reveals hierarchical and epistatic genetic control of flavonoid metabolism in soybean

Metabolic phenotypes are often governed by complex genetic architectures involving both additive and non-additive effects. However, the extent to which epistatic interactions contribute to the pathway-level regulation of plant metabolism remains unclear. In this study, we investigated the genetic architecture of flavonoid-related metabolites using metabolomic and genomic data from 200 soybean accessions cultivated under multiple environmental conditions. Broad-sense heritability estimates revealed that many metabolites were under strong genetic control, particularly flavonoid-related metabolites. Principal component analysis-based metabolome-wide genome-wide association studies identified four major loci associated with flavonoid metabolic variation, including a locus corresponding to flavonoid 3'-hydroxylase. Conditional analyses based on multilocus genetic backgrounds demonstrated that the effects of downstream loci were highly dependent on upstream genotypes. In particular, single-nucleotide polymorphism effects were frequently detectable only in specific allelic backgrounds defined by the major flavonoid 3'-hydroxylase locus, consistent with strong epistatic interactions among loci. Bayesian network analyses further supported a hierarchical genetic structure consistent with upstream regulation of downstream loci across the flavonoid biosynthetic pathway. These results demonstrate that highly heritable metabolic phenotypes can be controlled by a few loci exhibiting both additive and context-dependent non-additive effects. Our findings provide evidence that pathway-level metabolic diversity in soybean is generated through hierarchical and epistatic genetic control involving a limited set of key loci.

plant biology↗
bioRxiv2026

Cross Potential Selection for Multiple Traits Considering the Progeny Distribution of Future Inbred Lines in Plant Breeding Programs

In plant breeding, it is often necessary to improve a target trait while maintaining other essential traits within desirable ranges. When genetic relationships exist among these traits, improvements in the target trait may lead to undesirable changes in essential traits, complicating cross selections. In such cases, it is critical to select cross-pairs that are expected to produce progeny that satisfy the requirements for all traits. The progeny distribution of each crossing pair can be predicted using the estimated genotypic values and genetic (co)variances of the target and essential traits. By utilizing this distribution, the probability of generating progeny that satisfy predefined trait requirements can be evaluated, allowing a direct comparison of alternative crosses. In this study, we developed Cross Potential Selection for Multiple Traits (CPS-MT), a breeding strategy designed to improve a target trait while maintaining one or more essential traits within desirable ranges. CPS-MT extends the original Cross Potential Selection (CPS) framework to explicitly handle trade-offs between traits under genetic correlations. We evaluated the performance of CPS-MT through simulations involving four types of genetic relationships and two genetic causal factors between traits, resulting in seven scenarios. Across all scenarios, CPS-MT consistently improved the likelihood of obtaining desirable progeny, indicating that CPS-MT provides a practical and effective framework for cross selection under multi-trait constraints in breeding programs. Article SummaryThis study developed Cross Potential Selection for Multiple Traits (CPS-MT), a new breeding strategy designed to improve a target trait while maintaining one or more essential traits within desirable ranges. CPS-MT evaluates crossing pairs by predicting progeny distributions based on estimated genotypic values and genetic covariances, enabling direct comparison of alternative crosses under multi-trait constraints. Through simulations incorporating four types of genetic relationships and two causal factors (seven scenarios), CPS-MT consistently increased the likelihood of obtaining progeny that satisfied the predefined trait requirement. These results indicate that CPS-MT provides a practical, robust framework for target trait improvement under trait constraints.

genetics↗
bioRxiv2026

Efficient Optimization of Genotype Pairs for Intercropping using Genomic Prediction and Bayesian Optimization

Intercropping is a promising strategy to improve productivity and sustainability in agricultural systems, but designing effective genotype combinations remains a major challenge owing to the rapid increase in possible pairings as the number of candidate genotypes increases. This creates a practical bottleneck because field evaluation of all combinations is infeasible under realistic resource constraints. Here, we propose a framework that integrates genomic prediction and Bayesian optimization to support efficient decision-making for intercropping system design. Using genome-wide marker data from sorghum and soybean, we simulated intercropping performance across 5,214 genotype pairs under certain genetic architectures, including variation in heritability, correlations between direct and indirect genetic effects, and the contribution of pair-specific interactions. Genomic prediction models incorporating direct and indirect genetic effects substantially improved prediction accuracy compared with models based on direct genetic effects alone, and inclusion of specific mixing ability further enhanced the performance under high-heritability conditions. When coupled with Bayesian optimization, the models rapidly identified superior genotype pairs, requiring fewer evaluation cycles than random or prediction-only search strategies. Acquisition functions that account for predicted uncertainty were most effective in complex scenarios involving interaction effects or negative correlations between direct and indirect effects. These results demonstrate that combining genomic prediction with Bayesian optimization can substantially reduce the experimental burden associated with intercropping design, while improving the efficiency of identifying high-performing genotype pairs. The proposed framework provides a practical approach for prioritizing candidate mixtures in breeding and field evaluation, and contributes to the development of data-driven strategies for sustainable agricultural systems. HighlightsO_LIA data-driven framework was developed to optimize genotype pairs in intercropping. C_LIO_LIModeling indirect effects improved prediction accuracy across genotype pairs. C_LIO_LIPair-specific interactions enhanced prediction under high-heritability conditions. C_LIO_LIBayesian optimization identified superior pairs under limited evaluation capacity. C_LIO_LIThe framework reduces field-testing requirements for intercropping system design. C_LI

genomics↗
bioRxiv2025

Leveraging species-wide variation and patterns of adaptation to inform pecan crop improvement efforts

The genetic basis of adaptation is a fundamental question in evolutionary biology, and understanding how species will be able to adapt to changing conditions across their range has important implications for conservation and agriculture. To accurately interrogate the genetics of adaptation and assess the adaptive capacity of a species requires also characterizing the ways other mechanisms, including geographic distance and population dynamics, shape genetic variation. Pecan is an ecologically, culturally, and economically important North American tree, and a broader understanding of the genetics of environment adaptation will aid pecan conservation, breeding, and commercial management. Here, we use an expansive set of more than 700 pecan genotypes in combination with the first haplotype-resolved genome assembly for pecan to assess species-wide genetic variation and evaluate environmental adaptation across the native distribution. We identify five gene pools in pecan, with the lowest diversity in southern gene pools, and present evidence that gene pools began differentiating during multiple glacial cycles. Using complementary genotype-environment association approaches, we infer species-wide patterns of environmental adaptation. With these results, we predict mismatches in adaptation for pecan genotypes to different environments, including future environment scenarios. We see that in all locations, present-day genotypes incur some level of predicted maladaptation to simulated future environments, but current genetic diversity may provide a valuable source of resilience to future conditions through assisted migration. These results expand the understanding of environmental adaptation in pecan and provide insight into how long-lived species will be able to adapt to future conditions.

evolutionary biology↗
bioRxiv2025

Reciprocal BLUP: A Predictability-Guided Multi-Omics Framework for Plant Phenotype Prediction

Sustainable improvement of crop performance requires integrative approaches that link genomic variation to phenotypic expression through intermediate molecular layers. Here, we present Reciprocal Best Linear Unbiased Prediction (Reciprocal BLUP), a predictability-guided multi-omics framework that quantifies cross-layer relationships among the genome, metabolome, and microbiome to enhance phenotype prediction. Using a panel of 198 soybean accessions grown under well-watered and drought conditions, we first evaluated four direction-specific prediction models (genome[->]microbiome, genome[->]metabolome, metabolome[->]microbiome, and microbiome[->]metabolome) to estimate the predictability of individual omics features. Then, we evaluated whether subsets of features with high cross-omics predictability improve phenotype prediction. These cross-layer models identify features that play physiologically meaningful roles within multi-omics systems, enabling us to prioritize variables that capture coherent biological signals enriched for phenotype-relevant information. As a result, metabolome features were highly predictable from microbiome data, whereas microbiome predictability from metabolomic data was weaker and more environment-dependent, revealing an asymmetric relationship between these layers. In the subsequent phenotype prediction analysis, the model incorporating predictability-based feature selection substantially outperformed models using randomly selected features and also achieved prediction accuracies comparable to those of the full-feature model. Under drought, phenotype prediction models based on metabolomic or microbiomic kernels (MetBLUP or MicroBLUP) outperformed the genomic baseline (GBLUP) for several biomass-related traits, indicating that environment-responsive omics layers captured phenotypic variation not explained by additive genetic effects. Our results highlight hierarchical interactions among genomic, metabolic, and microbial systems, with the metabolome functioning as an integrative mediator linking genotype, environment, and microbiome composition. The Reciprocal BLUP framework provides a biologically interpretable and practical approach for integrating multi-omics data, improving phenotype prediction, and guiding omics-based feature selection in plant breeding.

bioinformatics↗
bioRxiv2025

Progeny-based genomic selection reveals untapped genetic potential in an underutilized medicinal plant, Perilla frutescens

O_LIDespite their substantial therapeutic value, medicinal plants have undergone limited genetic improvement through breeding because of the scarcity of expert breeders. Moreover, quantifying bioactive compounds is expensive. Genomic selection (GS), which leverages genome-wide markers to predict breeding values and assemble favorable alleles, offers a practical way to unlock latent genetic potential. As a model case, we evaluated GS in red perilla (Perilla frutescens). C_LIO_LIBuilding on previous work, we implemented a cross-selection strategy that prioritized segregation variance by selecting crosses based on predicted additive genotypic values of the progeny, and evaluated its effectiveness through actual crossing experiments targeting three key medicinal compounds. C_LIO_LIProgeny from GS-based crosses (Crs1-Crs7) outperformed those from phenotypic selection (Crs8) in the G2 generation, demonstrating a higher mean, greater variance, and superior top individuals. The best G2 individual exhibited nearly twofold higher levels of two target compounds relative to the existing cultivar Sekiho. C_LIO_LIThis study provides the first empirical demonstration that GS can improve multiple medicinal compounds in red perilla and highlights the effectiveness of cross-selection based on predicted progeny performance. In addition, the evidence presented here supports the broader application of GS in underutilized medicinal plants. C_LI

genetics↗
bioRxiv2025

Optimizing progeny allocation strategies in breeding schemes while updating genomic prediction models

Genomic selection has revolutionized breeding by enabling the early identification of superior individuals using genome-wide markers, enhancing breeding efficiency and accelerating variety development. Over the past decade, new selection and mating strategies -- leveraging optimization methods and other approaches -- have been introduced to improve various decision-making processes in breeding programs. However, optimizing breeding remains challenging when the positions and effects of quantitative trait loci are unknown. We developed a framework that optimizes breeding strategies while updating genomic prediction models during breeding schemes. By implementing intermediate model updates, we enabled re-optimization of allocation strategies based on updated predictions. Our simulations compared this approach with equal allocation and optimal cross selection methods across various selection intensities and genetic architectures. Results demonstrated our optimized allocation strategy significantly outperformed the other approaches under moderate to low selection intensities, particularly when combined with model updates. While genetic gains plateaued without updates, our approach enabled continuous improvement through the final generation. The framework showed exceptional robustness across different simulation conditions and better maintained genetic diversity while controlling changes in population structure. This confirms that optimized allocation strategies remain effective when using estimated marker effects rather than true effects, providing a practical framework for improving real-world breeding programs.

genomics↗
bioRxiv2025

Interpretable Machine Learning on Soybean Multi Omics Data Reveals Drought-Driven Shifts of Plant-Microbe Interactions

BackgroundPlant-microbe interactions in the rhizosphere are central to plant growth, nutrient acquisition, and stress resilience. Although multi-omics approaches enable comprehensive profiling of different biological layers, integrating these data to understand the mechanisms underlying plant-microbe symbiosis, particularly under drought stress, remains a challenge. ResultsGenomic, metabolomic, and microbiome data from 198 soybean accessions grown under both control and drought conditions were integrated to identify environment-specific predictive features of the plant phenotypes. We compared best linear unbiased prediction (BLUP), genome-wide association study (GWAS), and a nonlinear machine learning model to evaluate their ability to detect informative features. The machine learning models provided flexible variable selection and outperformed linear models in capturing nonlinear dependencies. Model interpretation using SHapley Additive exPlanations (SHAP) indicated that the isoflavone derivative, daidzin, and the drought-tolerant Candidatus Nitrosocosmicus, were major contributors to phenotypic variation, specifically under drought stress. SHAP-based interaction networks indicated cross-omics links, including connections between daidzin, gamma-aminobutyric acid (GABA), and Paenibacillus. ConclusionThe proposed interpretable machine learning approach for plant phenotype prediction identified multi-omics biomarkers and interactions, providing insights into plant adaptation to drought stress through environment-dependent rhizosphere networks and symbiotic associations.

bioinformatics↗
bioRxiv2025

Ridge Redundancy Analysis for High-Dimensional Omics Data

MotivationRedundancy Analysis (RDA) is a popular reduced rank regression approach for modeling the relationships between two sets of variables. In omics studies, RDA provides a flexible framework for investigating associations between high-dimensional molecular data. However, omics datasets often exhibit strong multicollinearity and suffer from the "large p, small n" problem, where the number of predictor variables p (features) exceeds the number of samples n. Ridge RDA addresses multicollinearity by introducing a ridge penalty in addition to the rank restriction. Despite these advantages, its application to high-dimensional omics data remains challenging due to i) difficulties in manipulating the large coefficient matrix of dimensions p x q, where q represents the number of response variables, and ii) the need to jointly select the rank and the regularization parameter that both influence the performance of the method. ResultsWe propose an efficient computational framework for ridge RDA that overcomes these challenges by leveraging the Singular Value Decomposition of the predictor matrix X. This approach eliminates the need for direct covariance matrix inversion, improving computational efficiency. Furthermore, we introduce a novel strategy that reduces the memory burden associated with the storage of the coefficient matrix from pq to (p + q)r, with r the chosen reduced-rank dimensionality. Our method enables an efficient grid search to jointly select the ridge penalty{lambda} and r via cross-validation. The proposed framework is implemented in the R package rrda, providing a practical and scalable solution for the analysis of high-dimensional omics data. AvailabilityThe R package rrda is available on CRAN: https://CRAN.R-project.org/package=rrda. Scripts and data used for the analysis are available on GitHub: https://github.com/Yoska393/rrda.

bioinformatics↗
bioRxiv2025

Haplotype Bias Detection Using Pedigree-Based Transmission Simulation: Traces of Selection That Occurred in Apple Breeding

With the increasing ability to integrate pedigree and genomic data, it is essential to evaluate their potential to uncover valuable genetic insights that can drive the advancement of crop breeding and conservation of genetic diversity. Pedigree analysis remains a fundamental approach for investigating the inheritance of phenotypic traits, exploring evolutionary history, and understanding hybridization processes in crop plants. Among these approaches, gene drop simulations using pedigree and allele origin data enable the construction of genetic maps and provide insights into complex genetic backgrounds. In this study, we developed a new method to identify useful genetic regions associated with single-nucleotide polymorphism (SNP) markers based on gene drop simulations, focusing on 185 Japanese domestic apple cultivars. By performing 10 million gene drop simulations, we generated null distributions for each founder haplotype, which revealed SNP markers with significant frequency biases, which is a potential signal for selection. Frequency biases were identified in eight founder haplotypes that were particularly consistent with genome-wide association studies peaks associated with key fruit traits such as malic acid and fructose content. Gene Ontology enrichment analysis suggested that these SNPs are not only associated with fruit traits but may also play a role in critical biological functions, including stress tolerance and reproductive processes, highlighting their broader relevance to crop resilience. Our integrative approach, which combines founder haplotype analysis with extensive gene drop simulations, effectively detects selection pressure, provides new insights into the genetic basis of apple breeding, and identifies SNP markers with strong potential to improve breeding programs.

genetics↗
bioRxiv2025

Integration of Proxy Intermediate Omics traits into a Nonlinear Two-Step model for accurate phenotypic prediction

Intermediate omics traits, which mediate the effects of genetic variation on phenotypic traits, are increasingly recognised as valuable components of genetic evaluation. In particular, rhizosphere microbiota play a crucial role in plant health and productivity; however, their complex interactions with host genetics remain challenging to model. Although two-step modeling frameworks have been proposed to integrate intermediate omics traits into phenotype prediction, existing approaches do not incorporate nonlinear relationships between different omics layers. To address this, we have proposed a two-step phenotype prediction framework that integrates genomic, rhizosphere microbiome, and metabolome (meta-metabolome) data, while explicitly capturing omicsomics nonlinearities. The first step is to predict meta-metabolome traits from genetic and microbial features, thus effectively isolating them from the environmental noise. In this process, intermediate "proxy" omics traits are generated as general biological information to provide robust models. The second step utilises this "proxy" to enhance the accuracy of the phenotype prediction. We compared the linear model (Best Linear Unbiased Prediction, BLUP) and the nonlinear model (Random Forest, RF) at each step, as demonstrated through simulations and empirical analysis of a multi-omics soybean dataset in which nonlinear modeling captures intricate omics interactions. Notably, our approach enables phenotype prediction without requiring the original meta-metabolome data used in model training, thereby reducing reliance on costly omics measurements. This framework integrates intermediate omics traits into genomic prediction to improve prediction accuracy and provide solutions for deeper insights into plant-microbiome interactions.

bioinformatics↗
bioRxiv2025

Optimization of crossing strategy based on the usefulness criterion in inter-population crosses considering different genetic effects among populations

In the breeding programs of self-pollinated plants, achieving genetic improvement in multiple traits can be challenging when relying solely on a single biparental population. Interpopulation crosses are employed to integrate favorable alleles from multiple biparental populations to overcome this limitation. In this context, it is crucial to consider the distinct genetic effects in different populations. In this study, we utilized a selection method based on the usefulness criterion (UC) to identify cross pairs suitable for interpopulation crosses. We expanded this approach to enhance breeding programs accounting for varying genetic backgrounds within the genomic selection framework. Using the medicinal plant red perilla as the study material, we conducted simulations to compare the efficacy of selection based on estimated genomic breeding values with that based on UC. Our findings demonstrate that the proposed method is effective in facilitating the simultaneous improvement of multiple traits, particularly by considerably increasing genetic gains among the top-performing individuals in the population. Furthermore, we provide guidelines for implementing interpopulation crosses, including recommendations for the optimal generation for crossing and the appropriate reference generation for calculating the UC. The results obtained in this study offer valuable insights for small-scale breeding programs aimed at simultaneously enhancing multiple traits through inter-population crosses and are applicable to a wide range of crops, including neglected and underutilized species. Key MessageHerein, a method has been proposed for selecting optimal cross pairs based on the genetic potential of progeny in inter-population crosses, considering different genetic effects among populations.

bioinformatics↗
bioRxiv2025

Estimating fruit tree growth curves in breeding field using fragmented longitudinal data: An application to citrus hybrid seedlings

Vegetative and reproductive growth in fruit trees is interconnected, and analyzing this relationship can provide valuable insights into fruit quality. However, characterizing vegetative growth through growth models is challenging because of the difficulty in obtaining longitudinal data, given the slow growth rate. In breeding fields, in contrast, seedlings of different ages are planted, allowing for simultaneous measurements that yield a dataset resembling longitudinal data with missing values --termed "fragmented longitudinal data." Because longitudinal data are obtained from a single measurement, they can potentially shorten the period required for growth curve estimation. Bayesian nonlinear models offer advantages in estimating curves from incomplete data. In this study, we generated fragmented longitudinal data using genome data with 45,929 markers from 624 citrus hybrid seedlings and applied a Bayesian nonlinear model to explore its potential. We also incorporated genomic information into the model to assess the impact of the estimation accuracy. Our simulations indicated that the Bayesian nonlinear models ability to interpolate missing values significantly improved the estimation performance. At best, the mean square error of the parameter characterizing the later growth stage was reduced by 84.3 mm2. Although the improvement from incorporating genomic information was modest, it still surpassed models that lacked genomic data. We also predicted the curves of untested individuals using the estimated parameters. Although the prediction accuracy of each parameter measured by the correlation coefficient was lower than 0.5, one parameter consistently showed a better accuracy. Further research is required to reveal the advantages of integrating genomic data for better predictions.

genetics↗
bioRxiv2024

Differentiable breeding: Automatic differentiation enables efficient gradient-based optimization of breeding strategies

MotivationDifferentiable programming frameworks like PyTorch and JAX revolutionized biological modeling. A foremost merit is that multiple components programmed separately can be put together so that the parameters are jointly optimized. Despite its proven value in agricultural applications, existing breeding simulators are non-differentiable, hindering integration into general deep learning systems. ResultsIn this paper, we present DiffBreed, a differentiable breeding simulator. Its performance was evaluated in gradient-based optimization of a progeny allocation strategy that maximizes the genetic gain. By utilizing gradient-based optimization, DiffBreed refined progeny allocation strategies, achieving superior genetic gains compared to a non-optimized equal allocation approach. These findings highlight DiffBreeds capacity to properly calculate gradient information through automatic differentiation. With its innovative design, DiffBreed is expected to transform future breeding optimization by integrating into modern deep learning workflows. AvailabilityThe proposed utomatic differentiation framework was implemented as the Python module "DiffBreed". This module and all the scripts used in this study, including the gradient-based optimization, are available from the "KosukeHamazaki/GORA-PT" repository on GitHub, https://github.com/KosukeHamazaki/GORA-PT. While the simulated datasets in the present study are available from the same repository, the optimized results by PyTorch were not shared in this repository due to the file sizes. Instead, all datasets, including the optimized ones, will be shared in the repository on Zenodo, https://doi.org/10.5281/zenodo.14046522.

bioinformatics↗
bioRxiv2024

ConvCGP: A Convolutional Neural Network to Predict Genotypic Values of Rice Traits from Compressed Genome-Wide Polymorphisms

The growing size of genome-wide polymorphism data in animal and plant breeding has raised concerns regarding computational load and time, particularly when predicting genotypic values for target traits using genomic prediction. Although several deep learning and conventional methods, including dimensionality reduction techniques, such as principal component analysis (PCA) and autoencoders, have been proposed to address these challenges by selecting subsets of polymorphisms or compressing high-dimensional data for predictive analysis. However, these methods are often computationally intensive and time-consuming. A major challenge in applying high-dimensional genomic data directly to deep-learning models is the substantial computational cost and time required for hyperparameter tuning and model training. To address these limitations, we propose a novel deep learning approach that combines convolutional neural networks (CNNs) to predict the genotypic data of target traits with autoencoders to compress high-dimensional genome-wide polymorphism data. We tested this framework on high-dimensional rice datasets, focusing on agronomic trait prediction. By combining CNNs with autoencoders, our framework outperformed other machine-learning methods and recently proposed compression methods, demonstrating its potential to efficiently address the computational challenges associated with high-dimensional genomic data.

bioinformatics↗
bioRxiv2024

Quantifying Forest Biomass and Genetic Contribution using Light Detection and Ranging

The growing focus on the role of forests in carbon sequestration highlights the importance of accurately and efficiently measuring biophysical traits, such as diameter at breast height (DBH) and tree height. Understanding genetic contributions to trait variation is crucial for enhancing carbon storage through genetic improvement of forest trees. Light detection and ranging (LiDAR) has been used to estimate DBH and tree height; however, few studies have explored the heritability of these traits or assessed the accuracy of biomass increment selections based on these traits. Therefore, this study aimed to leverage LiDAR to measure DBH and tree height, estimate tree heritability, and evaluate the accuracy of timber volume selections based on these traits using 60-year-old larch as the study material. Unmanned aerial vehicle (UAV) and backpack LiDAR were compared against hand-measured values. The accuracy of DBH estimations using backpack LiDAR resulted in a root mean square error (RMSE) of 2.7 cm and a coefficient of determination of 0.67. Conversely, the accuracy achieved with UAV LiDAR was 4.0 cm in RMSE and a 0.24 coefficient of determination. The heritability of DBH was found to be higher for backpack LiDAR than for UAV LiDAR and even exceeded that of hand measurements. Comparisons of the accuracy of timber volume selections based on the measured traits demonstrated comparable performances between the backpack and UAV LiDAR. Overall, these findings underscore the potential of using LiDAR remote sensing to quantitatively measure forest tree biomass and facilitate their genetic improvement of carbon-sequestration ability based on these measurements.

genetics↗
bioRxiv2024

High-throughput Phenotyping of Soybean Biomass: Conventional Trait Estimation and Novel Latent Feature Extraction Using UAV Remote Sensing and Deep Learning Models

High throughput phenotyping serves as a framework to reduce chronological costs and accelerate breeding cycles. In this study, we developed models to estimate the phenotypes of biomass-related traits in soybean (Glycine max) using unmanned aerial vehicle (UAV) remote sensing and deep learning models. In 2018, a field experiment was conducted using 198 soybean germplasm accessions with known whole genome sequences under two irrigation conditions: drought and control. We used a convolutional neural network (CNN) as a model to estimate the phenotypic values of five conventional biomass-related traits: dry weight, main stem length, numbers of nodes and branches, and plant height. We utilized manually measured phenotypes of conventional traits along with RGB images and digital surface models from UAV remote sensing to train our CNN models. The accuracy of the developed models was assessed through 10-fold cross-validation, which demonstrated their ability to accurately estimate the phenotypes of all conventional traits simultaneously. Deep learning enabled us to extract features that exhibited strong correlations with the output (i.e., phenotypes of the target traits) and accurately estimate the values of the features from the input data. We considered the extracted low-dimensional features as phenotypes in the latent space and attempted to annotate them based on the phenotypes of conventional traits. Furthermore, we validated whether these low-dimensional latent features were genetically controlled by assessing the accuracy of genomic predictions. The results revealed the potential utility of these low-dimensional latent features in actual breeding scenarios.

plant biology↗
bioRxiv2024

Cross Potential Selection: A Proposal for Optimizing Crossing Combinations in Recurrent Selection Based on the Ability of Future Inbred Lines

In plant breeding programs, rapid production of novel varieties is highly desirable. Genomic selection allows the selection of superior individuals based on genomic estimated breeding values. However, it is worth noting that superior individuals may not always be superior parents. The choice of the crossing pair significantly influences the genotypic value of the resulting progeny. This study introduced a new strategy for selecting crossing pairs, termed Cross Potential Selection (CPS), designed to expedite the production of novel varieties. The CPS assesses the potential of each crossing pair to generate a novel variety. It considers the segregation of each crossing pair and computes the expected genotypic values of the topperforming individuals, assuming that the progeny distribution of genotypic values follows a normal distribution. We simulated a 10-year breeding program to compare CPS with three other selection strategies. CPS consistently demonstrated the highest genetic improvements among the four strategies in early cycles. In particular, during the middle cycles of the breeding program, CPS exhibited the highest genetic improvement of 73% of the 300 independent breeding simulations. In a long-term breeding scheme, some progeny distributions of genotypic values may deviate from normal distribution, affecting the efficiency of CPS. Nevertheless, compared with the other three strategies, CPS achieved significant short-term genetic improvements. In conclusion, CPS holds substantial promise for enhancing the efficiency of plant breeding programs. Article SummaryThis study introduces a novel plant breeding strategy termed Cross Potential Selection (CPS), which was designed to expedite the production of novel varieties. The CPS evaluates the potential of each crossing pair for the target generation. Through comparative breeding simulations, CPS demonstrated superior performance over the three alternative breeding strategies, particularly in the early cycles. These findings suggest that CPS holds significant promise for enhancing plant breeding efficiency.

genetics↗