Theoretical and Applied Genetics

Sudden Death Syndrome (SDS), caused by Fusarium virguliforme, can significantly impact soybean yields. To advance the understanding of SDS resistance in plant introductions, we conducted genome-wide association studies (GWAS) on a set of 351 accessions from the USDA Soybean Germplasm Collection. The analysis revealed seven single nucleotide polymorphisms (SNPs) significantly associated with foliar SDS resistance; three significant QTL (on chromosomes 3, 9 and 11) were detected with Fixed and Random Model Circulating Probability Unification (FarmCPU) and four (on chromosomes 5, 6 and two on chromosome 7) with Bayesian-information and Linkage disequilibrium Iterative Nested Keyway (BLINK). Localized linkage disequilibrium blocks were estimated for these regions. Notably, all of them overlapped with regions where QTL associated with SDS resistance had been previously reported and where within, or close to, sequences annotated as implicated in plant defense mechanisms, such as toxin transporters and transcription factors. The architecture of SDS resistance appears quantitative and polygenic, with most loci explaining modest phenotypic variance. Six plant introductions in this study--PI 175188, PI 307889A, PI 323559, PI 578307C, PI 445683, and PI 323558--harbor favorable alleles at all seven loci, providing high-value resources for resistance introgression. This study underscores the value of leveraging the USDA Soybean Germplasm Collection for the discovery of resistance loci and provides a foundation for marker-assisted breeding to enhance SDS resistance in elite soybean lines.

Breeding programmes choose genomic prediction models by accuracy rankings, and the metric behind a ranking decides which model wins. Selecting the best genotypes within a trial depends only on the order of the predictions, so a metric that guides the choice must be unchanged by any order-preserving change of the predictions and respond when their order is reversed. Of the 22 official metrics of the Genomes to Fields maize competition, one meets both conditions, the within-environment rank correlation; the within-environment Pearson correlation is protected against rescaling only; and twenty fail, including the root mean squared error that decided the 2022 ranking. Switching between the competitions two official metrics reverses 27 % of team pairs, and an exact decomposition shows why pooled metrics mislead: in three constructed panels, 94-95 % of the attainable square of pooled correlation is carried by the environment means. The accuracy difference a trial can resolve falls as about 3/{surd}N with its N genotype-environment cells: at least 0.03 at the competitions design and 0.04-0.12 once calibrated on real panels, against 0.003 between the first two teams; tests that two correlations differ declare far smaller differences, which do not reliably change selection. In soybean, where the two most accurate methods stand clear of most others by more than the trial resolves, ranking by either within-environment correlation selects better material, 98-99 % of the attainable selection gain in held-out environments against 13-34 % for error-magnitude rankings; in common bean and a Chinese maize population, where the leading methods lie within the resolvable gap, all rules pick material of similar value. GPverdict runs these checks on any programmes cross-validated predictions.

Genotype-by-environment interactions can complicate plant breeding because genetic effects may vary across weather regimes, locations, and years. Although genomic prediction models can use dense marker data to predict complex crop traits, they often provide limited information about how marker relevance changes across environmental conditions. This study introduces a graph-based framework for estimating environment-associated marker rankings from multi-environment crop trial data. We introduce an attention-based graph neural network that jointly represents SNP markers and weather-based environment vectors as interacting nodes and is designed to learn environment-associated marker rankings from supervised yield data. The model is trained using yield prediction as the supervised objective, and the learned marker-level attention signals are aggregated to estimate marker rankings overall and within groups of similar environments. We evaluate the framework on two genotype-by-environment datasets. In both datasets, we examine whether learned marker rankings are stable across independent training runs and determine the overlap between high-ranked markers and external GWAS references. In a maize dataset with data from 212 different site-years, we rank markers separately within each weather-defined environment cluster and define a global marker set as the markers shared across all clusters. This provides a structured way to examine how marker rankings vary across weather-defined groups and which markers remain important across environments. We further encode each weather variable with a separate LSTM combined by attention, which makes the contribution of individual weather variables readable, and compare the resulting variable ranking against an independent SHAP-based attribution. Genes located near the top-ranked markers are also examined for functional enrichment. Overall, the results suggest that graph-based, environment-conditioned marker attribution may complement genomic prediction as an exploratory tool for studying context-dependent genomic signals in multi-environment trials.

The mechanisation of agriculture has achieved major increases in labour productivity but has also led to an erosion of agrobiodiversity. Neglected and underutilized crops such as common buckwheat (Fagopyrum esculentum Moench) offer the potential to diversify food production systems but require major genetic improvement to compete with established crops. Although target traits for breeding such as more homogeneous ripening, reduced height, larger seeds and higher and more stable yields are well known, large-scale and field-based phenotyping data of diverse common buckwheat accessions across multiple environments are still scarce. Here, we tested the agronomic performance of a common buckwheat germplasm collection consisting of 170 accessions of diverse origin. The collection was tested at seven year-sites, and 19 agronomic traits related to phenology, growth habit and yield were assessed. Most traits showed high repeatability (0.31 < R2 < 0.98) and heritability (0.45 < H2 < 0.97), indicating genetic control of these traits. Coefficients of variation ranged between 6.19% and 46.15%, showing the high phenotypic diversity within the collection. By combining multiple traits, distinct ideotypes for different roles in the crop rotation were investigated, and accessions with promising agronomic profiles identified. Genome-wide association studies revealed marker-trait associations for 18 traits, allowing the identification of putative candidate genes influencing these traits such as a putative GRF-INTERACTING FACTOR associated with flowering time and biomass production. Our results show that the collection offers vast phenotypic diversity that can be exploited for the development of improved varieties tailored to meet specific agronomic purposes. The efficient utilization of these genetic resources should be supported by the application of genomics-based breeding approaches. The development of genetic markers and search for putative candidate genes mark important steps towards the establishment of efficient breeding schemes. Thus, these genomic resources provide a foundation for accelerating the genetic improvement of common buckwheat and developing agronomically competitive varieties.

Caraway (Carum carvi L., 2n = 2x = 20) is a medicinal and aromatic plant of the Apiaceae family valued for its essential oil with the major terpenoids carvone and limonene. While the essential oil content represents an important quality trait, low seed shattering and high thousand-grain weight (TGW) are key determinants of seed yield. As molecular tools remain scarce within Apiaceae breeding beyond carrot, we conducted a QTL mapping for these seed-related traits using two previously established F2 populations (Effective N(P1) = 236; N(P2) = 266). In P1 with the non-shattering maternal parent Rekord, QTL mapping revealed at least one shattering locus on LG03 (PVE: 9.9%). In addition, loci associated with essential oil (LG08, 10.5% PVE) and TGW (LG09, 11.1% PVE) were detected. In P2, we identified six essential-oil loci (LG02-06, LG10; combined PVE: 49.2%), with five loci affecting both carvone and limonene, and one (LG03) exclusively limonene. Additionally, a major QTL for TGW mapped to LG02 (PVE: 15.6%). For stalk attachment rate one minor QTL was detected (LG07, 8.5% PVE), while no QTL co-localized with the major shattering locus. To pave the path towards fine-mapping, we designed and validated 26 diagnostic PACE markers covering five key QTLs.

Efficient yield prediction under field conditions remains a key challenge for crop breeders and agronomists. YieldLearn is an interpretable machine-learning pipeline developed to estimate yield-related traits in durum wheat (Triticum durum) using physiological indicators measured by the Hansatech Instruments Pocket PEA. The workflow integrates fluorescence parameters (Area, Fo, Fm, Fv/Fm, PI Inst.) with field-measured agronomic traits (Booting, Height, Maturity, Lodging, Yield kg/plot, and Adjusted Yield kg/ha). Five supervised learning algorithms multinomial logistic regression, random forest, support vector machine (radial kernel), k-nearest neighbors, and decision tree were evaluated using five-fold cross-validation. Models achieved average classification accuracy around 0.57 with AUC values exceeding 0.58. SHAP (SHapley Additive exPlanations) visualizations revealed that high Fv/Fm and PI Inst. values were positively associated with yield potential, whereas elevated Fo indicated stress-induced yield reduction. An interactive Shiny App enables real-time prediction and feature-importance exploration. Given the small sample size (165 plots) and absence of environmental covariates such as temperature or soil moisture, YieldLearn should be regarded as a reproducible exploratory framework rather than a production-ready predictor. Nonetheless, it offers a valuable foundation for guiding experimental design, validating physiological indicators, and supporting the next generation of scalable, interpretable crop-yield models.

Genomic selection (GS) and trait discovery in crops require reliable marker datasets, yet variant discovery remains constrained by the lack of standardised, high-confidence reference resources. Publicly available plant genomic resources are often based on outdated reference genomes, heterogenous pipelines, and generic variant hard-filtering thresholds derived from human studies. Consequently, implementation of the Genome Analysis Toolkit (GATK) Best Practices framework, including machine learning-based variant quality score recalibration (VQSR), remains challenging in crop genomes. Here, we present a crop-adapted framework for unsupervised variant quality assessment and VQSR enabling advanced SNP and insertion-deletion (INDEL) discovery without predefined high-confidence reference variant sets. Using Brassica napus as a model, we use the Illumina 60k Brassica SNP array to generate a reference-specific truth set for data pre-processing and independent evaluation of an unsupervised machine-learning approach for constructing cohort-specific VQSR resources. Gaussian mixture model (GMM)-based clustering of multidimensional variant annotation profiles identifies high-, intermediate-, and low-quality variant classes, providing flexible training resources for VQSR without requiring labelled training data. We further validate the framework in diploid Brassica oleracea and hexaploid wheat (Triticum aestivum) across cohorts differing in population size, genetic diversity, and genome architecture. Despite population- and species-specific differences in variant annotation profiles, GMM-based clustering consistently identifies informative variant quality classes and enables cohort-specific recalibration of both SNPs and INDELs. Our results demonstrate that unsupervised machine learning can extend GATK Best Practices to crop species lacking comprehensive genomic resources, providing a scalable approach for generating high-quality variant datasets for crop genomics and downstream breeding applications.

In common wheat (Triticum aestivum L.), variation at the Vrn1 and Ppd1 loci changes plant phenology, resulting in differential adaptation useful to breeders. Because the effects of phenology on grain yield are conditional on environmental factors, the incorporation of markers for Vrn1 and Ppd1 alleles into breeding approaches has proven difficult. Historic phenotypic, genotypic, and environmental data can provide insight into the relationship between major phenology alleles and grain yield as mediated by environmental variables. Analyses of eight years of breeding trials (1,038 lines at 219 site-years) determined that weak vrn1 alleles and Ppd1 insensitivity alleles have variable effects. These effects change depending on an environment's winter temperature and latitude, respectively, but the environmentally-driven changes in effect of vrn1 alleles are much more variable than effects of Ppd1 alleles. Mediation analyses showed phenologically-dependent environmental variables are one mechanism through which phenological changes created by vrn1 and Ppd1 variants alter yield across multiple developmental stages. Environmentally-driven relationships between flowering time and yield were modeled to estimate yield allele effects as a function of effects on phenology and environmental variables. Weak winter alleles at vrn1 may play a stabilizing role on yield, due to the correlation of winter conditions that drive larger effects with later-season warm temperatures that penalize later maturity. A better mechanistic understanding of how these loci drive environment-specific yield effects may facilitate utilization of markers for these alleles and improve predictive modeling of yield.

Forward genetics is a powerful approach for gene discovery, but identifying causal mutations becomes difficult when mutants are maintained in heterogeneous populations with uncertain pedigrees. This is exemplified by classical tasselseed (ts) mutants, which have long served as a genetic model for studying sex determination and carpel suppression. Decades of repeated outcrossing to diverse inbred lines have created substantial genetic heterogeneity, limiting the effectiveness of conventional bulked-segregant analysis sequencing (BSA-Seq). To address this, we developed a BSA-Seq framework that integrates flexible experimental designs, multiple reference genomes, and complementary statistical methods tailored for genetically heterogeneous populations. Applying this framework revealed that reference genome selection is critical for mapping success and that Euclidean distance raised to the fourth power (ED4) outperformed homozygosity mapping (HM). Furthermore, the framework enables simultaneous mapping of multiple mutations within a single population, eliminating the need for additional mapping populations. Applying this framework to 26 ts mutant stocks from the Maize Genetics Cooperation Stock Center, we successfully mapped 24 mutants to genomic intervals containing known ts genes, while the remaining mutants mapped to distinct genomic intervals, defining novel candidate regions underlying carpel suppression. Together, these results demonstrate that historical mutant collections represent an underutilized resource for gene discovery and establish a generalizable mapping strategy for unlocking their genetic potential across diverse species.

Amaranthus tricolor is a climate-resilient orphan crop whose betalain pigments contribute to nutritional quality and stress tolerance. However, breeding efforts have traditionally relied on field yield trials due to the limited availability of genomic information and genetic resources. To address this gap, we first used the Illumina and PacBio sequencing platforms to assemble a novel reference genome of A. tricolor, using the Japanese cultivar Bayam. By combining this reference genome with a core collection developed in our previous study, we aimed to identify genomic regions associated with pigmentation-related traits across winter and summer seasonal environments. Betalains are antioxidant phytonutrients that contribute to abiotic stress tolerance, making the regulation of their biosynthesis under heat stress particularly relevant for developing climate-resilient crops such as A. tricolor. As a result, ATR1.0CH03_5131103, which is associated with ATR1.0ch03g000565, was identified under the winter seasonal environment. Meanwhile, transcriptomic profiles in this study revealed that ATR1.0ch03g000565 was differentially expressed under both control and heat stress conditions, suggesting that it may be associated with the regulation of CYP76AD1 expression under optimal conditions. Under the elevated-temperature treatment, transcript levels of CYP76AD1 and 5GT decreased in Am087, coinciding with lower betacyanin-dominated absorbance indices. cDOPA-like homologs were upregulated; however, pigmentation-related absorbance remained reduced. Further molecular experiments, such as CRISPR-Cas9 knockouts or overexpression studies, are necessary to validate the regulatory mechanisms of CYP76AD1 and 5GT under heat stress.

Boron toxicity constrains grapevine (Vitis) production in irrigated arid and semi-arid regions, yet most available rootstocks are sensitive to excessive boron. Although underutilized Vitis species may harbor sources of adaptation, the genetic basis of natural variation in boron exclusion remains poorly understood, limiting the development of boron-tolerant rootstocks. We screened 267 wild Vitis accessions representing 16 species for their ability to limit leaf boron accumulation under excess boron conditions. Leaf boron concentration varied 190-fold among accessions and was highly heritable (entry-mean H^2 = 0.71). Its correlation with stomatal conductance was modest (r = 0.34), indicating that transpiration rate explained only a small proportion of the variation in leaf boron accumulation. Leaf boron was correlated with leaf chloride (r = 0.36), but their largely non-overlapping association profiles supported distinct genetic control. Genome-wide association analysis identified a suggestive association on chromosome 11. Independent quantitative trait locus mapping in a full-sib F1 population (n = 158) derived from Longii-9018 (V. acerifolia) x GRN3 (commercial rootstock) identified qBEx11.1 in the same syntenic region on chromosome 11, which explained 19.5% of the phenotypic variance. Positional candidate genes near the association peak included a pectinesterase and a MATE efflux transporter, suggesting roles for cell wall modification and membrane transport. Leaf spectral reflectance was associated with leaf boron concentration in both populations, particularly at visible and red-edge wavelengths, indicating potential for high-throughput phenotyping. Together, these findings identify qBEx11.1 as a priority locus for boron exclusion and support integrated genetic and spectral approaches for breeding boron-tolerant grapevine rootstocks.

Understanding the genetic architecture of ecophysiological traits related to carbon gain and water loss is crucial to develop cultivars tolerant to abiotic stresses and that maintain productivity in the context of climate change. This can be achieved both through experiments in the greenhouse on potted plants under controlled conditions and in the field. However, the stability of genetic determinism of ecophysiological traits across environmental conditions has rarely been evaluated. We studied a grapevine diversity panel grown both in a greenhouse and in the field, to investigate the impact of growing conditions on leaf ecophysiological traits related to leaf morphology, fluorescence, gas exchange and water use efficiency. The correlation structure between traits was conserved within each environment even though, for most traits, cultivars ranking was not conserved between both environments. However, a set of cultivars, sharing the same functioning pattern in terms of gas exchange and fluorescence in both environments, was identified. Single- and multi-environment Genome-Wide Association Studies (GWAS), based on a dense genotyping matrix, were used to dissect the genetic architecture of leaf traits. The Quantitative Trait Loci (QTLs) identified through single-environment analyses differed substantially from those detected using multi-environment ones. Multi-environment approaches highlighted shared QTLs across environments for gas exchange and fluorescence traits. Candidate genes associated with carbon- and water-related functions were further identified in these genomic regions. Despite the weak correlation between greenhouse and field trait values, combining both environments enabled the identification of robust genetic regions associated with ecophysiological traits of interest.

Petal coloration is a key ornamental trait in Cyclamen persicum Mill. and is closely associated with anthocyanin composition and relative accumulation, yet the major anthocyanins and candidate molecular factors associated with color variation remain unclear. In this study, anthocyanin profiling and transcriptome analysis were integrated to investigate three Cyclamen materials with distinct petal colors, BXK, FXK, and SHXK. Anthocyanin profiling showed clear separation among the three materials, indicating that petal color divergence was associated with differences in anthocyanin composition and relative abundance. Petunidin-, peonidin-, delphinidin-, malvidin-, and cyanidin-related derivatives exhibited distinct accumulation patterns. Transcriptome analysis revealed marked transcriptional divergence, with phenylpropanoid and flavonoid biosynthesis repeatedly enriched across comparisons, whereas isoflavonoid and flavone and flavonol biosynthesis were enriched in selected comparisons. Mapping differentially expressed genes onto the anthocyanin biosynthetic pathway revealed branch-specific expression patterns of structural genes, including CHS, F3H/F3'H, F3'5'H, DFR, and ANS. Notably, Cluster-26256.0, annotated as a putative anthocyanidin synthase gene, showed a progressive increase in expression from BXK to FXK and SHXK, consistent with increasing petal pigmentation. Several DFR-related transcripts also showed relatively high expression in SHXK. Co-expression analysis linked key anthocyanins, structural genes, and transcription factors, while qRT-PCR generally confirmed RNA-seq trends. These results suggest that late anthocyanin biosynthesis, particularly ANS-related transcriptional variation, may contribute to dark red petal coloration. This study provides a metabolite-transcript framework for Cyclamen petal coloration and identifies ANS-related genes as priority candidates for future functional studies and flower-color breeding.

We performed a screening to evaluate the responses to water deficit before the flowering stage in a panel of 31 genotypes composed of one commercial hybrid, one wild species, multiple ex-PVP (vintage) lines, several commercial heirlooms from the Northeast USA, Mexican landraces from hot and humid environments and genome edited lines with altered flowering and inflorescence traits. We characterized 18 different morphological, physiological and biochemical traits to assess the response to 40% water deficit. We observed diverse morphological, physiological and biochemical responses across the 31 genotypes screened, suggesting the presence of useful variation for breeding for drought tolerance in cultivated tomato. Interestingly, some heirlooms exhibited superior performance under drought, suggesting they can be used as donors to discover novel genes involved in drought responses. Unexpectedly, we found mutants in SlCLE9, a paralog of the stem cell regulator SlCLV3 known to co-regulate meristem proliferation in tomato, exhibited enhanced drought tolerance. Notably, the slcle9 mutant also conferred drought tolerance when used as rootstock with grafted scions from our reference cultivar M82, suggesting SlCLE9 influence root architecture and drought response. Consistent with this, we found SlCLE9 is transcriptionally upregulated in roots and leaves during drought stress, and exogenous application of SlCLE9 modifies root growth and response to drought. Altogether, our screening uncovered new potential trait donors for breeding against drought stress and describes a novel role for SlCLE9 in drought tolerance.

To reduce reliance on pesticides in potato production, alternative partial control methods are needed, yet their efficacy and reliability are seldom quantified over multiple years. Canopy architecture is a potential lever against late blight, caused by Phytophthora infestans, a pathogen whose infection and development are highly driven by humidity and temperature. However, its practical value under field conditions in pure crop stands remains poorly documented. We compared two commercial potato cultivars with contrasting canopy architectures but similar leaf tissue susceptibility to P. infestans, in replicated field trials conducted over six years (2011 2016) under irrigation, and over three of these years also without irrigation. Disease and canopy development (height, closure, leaf area, stem number) were monitored throughout the growing seasons and analysed using non linear growth models. Monalisa, an erect and ramose cultivar, consistently slowed epidemic progress relative to Bintje, a semi-erect and leafy cultivar. This effect was significant in every non irrigated trial but in only half of the irrigated ones. No single canopy trait explained this suppression across all years and conditions. However, canopy closure at the time of inoculation was the most consistent correlate, supporting the hypothesis of a microclimate mediated effect. Disease reduction seldom translated into a yield benefit, except under irrigation, where Monalisa occasionally out yielded Bintje. Our results confirm that canopy architecture can complement other levers within integrated late blight management. Nevertheless, its significant year to year variability and dependence on environmental conditions must be accounted for if it is to be deployed as a reliable control method.

High temperatures during grain filling stage are becoming increasingly frequent, compromising both grain and eating quality and thereby driving demand for heat-resilient cultivars. Such conditions are known to reduce the expression of granule-bound starch synthase I (GBSSI) and starch branching enzyme IIb (BEIIb), which are involved in starch biosynthesis, resulting in a decrease in amylose content and an increase in long-chain amylopectin. Thus, the present study introduced functional mutation in starch synthase IIa (SSIIa) that increases the proportion of short amylopectin chains to genetically compensate for the high-temperature-induced increase in amylopectin long chain. Rice lines carrying the ss2a mutation were grown at two locations with cooler (Akita) and warmer (Okayama) temperatures. Their grain traits, starch structure, and eating quality were compared. The ss2a mutant lines showed an increased proportion of short amylopectin chains as well as an increased apparent amylose content. Furthermore, these alterations in starch structure varied with the grain-filling temperature of the cultivation sites, ultimately affected eating quality. These results suggest that enriching short amylopectin chain via the ss2a mutation can counteract the increase in long amylopectin chain caused by high temperatures during grain filling, thereby maintaining a desirable starch structure and eating quality.

Breedbase ( is a comprehensive, web-based, open-source breeding management and analysis software for plant breeding that was introduced to address gaps with existing software for plant breeding data. Breedbase is continually improved and developed, with new features being added as required by the breeding community. Since its inception, the capabilities of Breedbase were improved to better serve the diverse data modeling needs of additional crops beyond our original ensemble of cassava, banana, yam, sweet potato, tomato, rice, and wheat. We describe here major updates that have been made in the past 4 years, including (1) improved handling of repetitive measurements; (2) a redesigned experimental treatments interface; (3) a more fine-grained user permission system; (4) improved trial quality control workflows; (5) intercropping support; (6) the addition of a decision meeting feature that helps decision makers make the best selection with the available data; (7) transgenics and tracking tools and finally, (8) an improved field layout tool which consolidates several trials in one field. These new features integrate seamlessly with the previous feature set of Breedbase - making it one of the most feature rich breeding management softwares available.

Bread wheat harbours low genetic diversity due to bottlenecks during speciation, domestication, and selection; wild relatives therefore provide critical reservoir for climate-resilient breeding. Aegilops mutica, one of the few self-incompatible diploids in the Aegilops-Triticum species complex, is a putative parent in the homoploid hybridisation event that formed the D-genome clade and is a proven donor of disease resistance, winter hardiness and grain zinc accumulation. Ae. mutica is globally endangered, with a reference-quality genome available for only a single Central Anatolian accession. Using HiFi sequencing, we generated genome assemblies from two accessions from opposite extremes of the species' range: western Turkiye and Armenia. The four haploid assemblies recovered ~93% complete BUSCOs and comprised ~81% repetitive sequence, predominantly LTR retrotransposons (~55%), closely matching the reference. Comparison against the reference identified 145,871-155,535 structural variants per haplotype, including an 8.4 Mb inversion on chromosome 6T. Between 96,860 and 105,472 structural variants segregated between haplotypes within individuals and classic domestication loci did not overlap with any large structural variants. These assemblies and structural variant catalogue extend genomic resources for Ae. mutica, providing a basis for targeted introgression into wheat and conservation of a species for which populations in the South Caucasus have severely declined.

Enhancing grain carotenoid concentration to increase yellowness and provitamin A content in sorghum (Sorghum bicolor [L.] Moench) is a major breeding objective across sub-Saharan Africa. Carotenoid breeding is constrained by costly post-harvest phenotyping and can be accelerated through genetic markers that capture functional variation in key carotenoid biosynthesis genes. Zeaxanthin epoxidase (ZEP) is a major gene controlling sorghum grain carotenoid accumulation, and SNP-based KASP markers facilitate selection of high-carotenoid grain, although quantitative variation remains among lines carrying the favorable allele. We hypothesized that additional functional variation within ZEP and other carotenoid biosynthesis pathway genes contributes to this variation and can be revealed by a pangenome-informed analysis. Using a 33-member pangenome reference, we characterized sequence and structural variation at ZEP and other key carotenoid biosynthesis genes and applied pangenome-derived genotyping to study marker diversity associated with carotenoid accumulation. Evaluation of ZEP in the pangenome reference revealed previously uncharacterized structural variation that is absent from the BTx623 primary reference genome. A pangenome-based association analysis identified genetic markers associated with the accumulation of carotenoids which were specific to particular pangenome reference members, including in RTx430 and SRN39, both yellow endosperm lines. Machine learning identified markers in ZEP, {beta}-OH, ZDS, and Z-ISO as the most predictive for all carotenoid traits, suggesting a multi-locus genetic architecture underlies the accumulation of carotenoids in sorghum grain. By integrating sorghum pangenomic resources with machine learning, this study establishes a framework for pangenome-accelerated trait discovery and identifies new genetic targets for carotenoid biofortification in sorghum.

Motivation: Optimum contribution selection (OCS) maximizes genetic gain under a constraint on relatedness. Genomic relationship matrices are dense, so the quadratic term costs O(n^3) operations, and existing OCS software assumes pedigree sparsity. Results: We replace the relationship matrix by its leading eigenpairs, computed by randomized SVD, and solve OCS in factorized form. On Norway spruce population (n=5,525), ranks 10--30 kept the exact objective within 1% of the optimum at 110--330-fold speedups; on the QTL-MAS 2010 (n=3,226), ranks up to 100 were needed (24-211 fold). The required rank rises as the diversity constraint loosens. Gain alone is a misleading accuracy check, since coancestry exceeded the optimum by up to 28%, so we choose the rank on the exact objective, which can be checked without a dense solve. The speedup comes from computing only the leading eigenpairs. We extend the factorization to robust OCS (18-fold). Availability and implementation: Julia implementation, and examples are available at https://github.com/jonhar97/fastocs under the MIT license (archived at Zenodo, DOI 10.5281/zenodo.23214804).