January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 0 Building capacity and resources for genomic prediction in forage grass and legume breeding Rabat, Morocco, December 10 to 12, 2024 Workshop organised by: In collaboration with: Crops to End Hunger is funded by: Crops to End Hunger Workshop Report January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 1 Contents Introduction..................................................................................................... 2 Abstracts and Short Bios .................................................................................. 4 1. The ILRI genebank: a window to the global tropical forage biodiversity available for breeding programs – Dr. Alemayehu Teressa Negawo, ILRI, Ethiopia 4 2. Overview of forage genetic diversity studies in FFD/ILRI- Dr. Meki Shehabu Muktar, ILRI, Ethiopia .................................................................................. 4 3. Genomics and genetics research on forage species at the Earlham Institute (EI)-Dr. Jose De Vega, Earlham Institute ..................................................... 6 4. Developing genomic tools for key underutilized forage species in Africa – Dr. Abel Teshome Gari, ILRI, Ethiopia ....................................................................... 7 5. Genetic Diversity and Genome-Wide Association Studies for the Important Agronomic Traits of Oat (Avena sativa L.). Ms. Lidiya Ashenafi, ILRI, Ethiopia ........................................................................................................................ 7 6. Use of Genomic Predictions at the Global Barley Breeding Program – Dr. Miguel Sanchez-Garcia ............................................................................................... 8 7. Genomic prediction in multi-environment and genotype-environment interactions models–Dr. Claudia Samantha Perea, Alliance Bioversity-CIAT ......... 9 8. Development of a mid-density SNP panel for genomic prediction in interspecific hybrids selection in Urochloa spp. – Dr. James Brett, – Earlham Institute ....................................................................................................................... 10 9. Genome-Wide Association Study and Genomic Prediction of Forage Biomass-Related Traits in Urochloa spp. – Dr. Haileslassie Gebremeskel, ILRI, Ethiopia ...................................................................................................................... 11 10. Genome-wide association study and genomic prediction for important traits in guinea grass (Megathyrsus maximus) – Dr. Kefyalew Negisho Bayissa, ILRI, Ethiopia .............................................................................................................. 12 11. Genomic Selection in Napier grass. – Dr. Zewdinesh Damtew Zigene, ILRI, Ethiopia ...................................................................................................................... 13 12. Genomic Prediction of Forage Biomass-Related Traits in Oats (Avena Sativa) and Lablab (Lablab purpureus). – Mr. Hailu Lire Wachamo, ILRI, Ethiopia 14 13. Progress on QTL mapping for spittlebug resistance in interspecific F1 biparental families of Urochloa spp. based on the RADseq sequence data. Dr. Paula Andrea Espitia, Alliance Bioversity-CIAT....................................................... 16 January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 2 Introduction As part of the Crops to End Hunger (CtEH) project, a training program was implemented at ICARDA and ILRI with the objective to accelerate genetic gain in fodder crops and forage grasses through the integration of genomic selection in breeding programs, and to build capacities of NARES researchers on its application. This book of abstracts compiles the insights and findings from the workshop on research activities related to the CtEH project, held in Rabat, Morocco, from December 10 to 12, 2024. The workshop brought together a diverse group of researchers, practitioners, and stakeholders to exchange information and foster collaboration in the vital field of forage research. After 15 months, the fellows trained at ILRI presented on the application of genomic tools in tropical forages. In addition, the project integrated experts from Earlham Institute in the UK and Alliance of Bioversity International and CIAT who collaborate on the application of genomic tools into the Tropical Forages breeding program. On the first day, participants engaged in discussions around the genetic diversity of tropical forages and the importance of genomic tools in breeding programs. The opening plenary highlighted the objectives of the CtEH project and its contributions to tropical forage research. The day continued with presentations from trainers from ILRI, who shared valuable insights on the ILRI genebank and ongoing genetic diversity studies. Participants from Earlham Institute presented on the group’s research on genomics and genetics on forage species and ICARDA presented on the use of genomic predictions at the global barley breeding program. The second day focused on the presentations of research fellows addressing Genome-Wide Association Studies (GWAS) and genomic prediction methodologies. The Alliance of Bioversity International and CIAT opened the session with a discussion on genomic prediction in multi-environment settings and advances on QTL mapping for pest tolerance in Urochloa spp, complemented by further research findings from Earlham Institute. The five fellows from Ethiopian NARES who were trained at ILRI presented how they had applied these genomic tools on different forage grasses such as Urochloa spp. (syn Brachiaria), Megathyrsus maximus (syn. Panicum maximum), Cenchrus purpureus (syn. Pennisetum purpureus), forage legumes such as Lablab (Lablab purpureus) and on oats (Avena sativa), a temperate forage cereal. Closing the workshop, participants had the opportunity to visit ICARDA’s research facilities. This hands- on experience connected theoretical knowledge with practical applications in the field of forage research. January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 3 Hereby the abstracts of all contributors in this volume are presented, reflecting their innovative research and collaborative efforts toward advancing forage breeding and genomics. We hope these insights will catalyse ongoing discussions and inspire new pathways for collaborative research and development in the future. Chris Jones and Rosa Noemi Jauregui, Principal Scientists. About the Organizers Chris Jones is program leader for feed and forage development, a multidisciplinary research program involving a team of plant molecular biologists, physiologists and geneticists, and animal nutrition scientists at International Livestock Research Institute (ILRI). His work is directed towards accelerating the genetic improvement of feed and forage species in support of livestock production in developing countries. He has a PhD from the University of Dundee and has researched all aspects of plant biotechnology from academic to highly commercially driven projects. He Joined ILRI in July 2015 from the New Zealand Crown Research Institute, AgResearch. Rosa Noemi Jauregui is the Tropical Forages Plant Breeder at the Alliance of Bioversity International and CIAT, which is part of the CGIAR consortium. Her breeding and research focus is on tropical grasses such as Urochloa humidicola, interspecific Urochloa and Megathyrsus maximus. The end goal of these breeding programs is to realize livelihood and environmental benefits through the improvement of tropical forages in Latin America and the Caribbean, Africa and Asia, with an emphasis on smallholder crop-livestock-tree systems. Rosa earned a BS in Biological Sciences from the University of Buenos Aires specializing in plant genetics; later she earned a Master´s degree in Horticulture & Agronomy at the University of Californa-Davis followed by a PhD in Plant Breeding at Texas A&M University in College Station, TX. She joined the private sector first, as Purity Lead in Monsanto’s Soybean Breeding Program and later, as the Breeding and Research Lead at GENTOS, a private forage seed company in Argentina. There, she led a research-based, diverse breeding program including temperate grasses, legumes, and other forage species through a strong network of local and international collaborators. January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 4 Abstracts and Short Bios 1. The ILRI genebank: a window to the global tropical forage biodiversity available for breeding programs – Dr. Alemayehu Teressa Negawo, ILRI, Ethiopia There are several genebanks across the globe holding diverse and invaluable crop genetic resources. These genetic resources are crucial global public goods that contribute to improving life on the planet and hold significant potential to tackle development challenges and contribute to improved food systems for better nutrition and diets. Over the last four decades, ILRI forage genebank has been conserving nearly 2,000 species, the world most diverse tropical forage collection in any genebank worldwide. The genebank has distributed several thousand samples reaching users in over hundred countries globally. The large proportion of the collection are wild or crop wild relatives with little information and understanding of the species. The recent advances in modern molecular tools were used to narrow this gap through generating information that enhances our knowledge of the collection for better use of the diverse genetic resources. In line with this, the ILRI forage genebank, in partnership with other forage diversity and breeding teams, has been developing genomic tools and field performance data for key forage species. These data were used not only to study diversity embedded in the collection but also to develop subsets and core collection. These subsets/core collections are being distributed to users by the genebank. There are also ongoing CGIAR genebanks collaborative multidisciplinary projects that aim to identify genebank accessions with low carbon emission potential and resistance traits against economically important pests and diseases. In summary, the developed cores/subsets can serve as valuable tools as a gateway into the global diversity of the various forage species. Breeders and researchers can utilize these resources to develop improved varieties that address production challenges in the face of climate change and growing global food security concerns. About Alemayehu Teressa Negawo Alemayehu Teressa Negawo is the Forage Genebank Manager at the International Livestock Research Institute (ILRI) with expertise in crop genetics and the application of advanced molecular tools in genebank collection management. Alemayehu's research focuses on studying genetic diversity, subsetting collections, assessing duplicates, genome-wide association study and field evaluation of genebank collections. He is interested in leveraging modern molecular tools to enhance the conservation and sustainable use of genetic resources maintained in genebanks. Additionally, he is interested in utilizing genomic and phenotyping tools for selection of high-yielding and climate-resilient genebank genotypes, exploring the role of genebanks in climate change adaptation and mitigation, managing and analyzing genetic data, and contributing to the development of future professionals in plant genetic resource conservation and utilization. Alemayehu holds PhD in Plant Biotechnology (Natural Science) and MSc in Plant Biotechnology (International Horticulture), both from Gottfried Wilhelm Leibniz Universität Hannover, Germany and BSc in Horticulture from Jimma University, Ethiopia. 2. Overview of forage genetic diversity studies in FFD/ILRI- Dr. Meki Shehabu Muktar, ILRI, Ethiopia Tropical forages are an important source of animal feed and are crucial in rehabilitation of degraded lands as most of the species can grow in degraded and marginalized lands. Despite their importance, tropical forage species are not well domesticated, most are still in their wild state, have received little attention by January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 5 the conventional breeding and genetic studies, and they have not benefited as much from the current rapidly advancing molecular technologies. The forage diversity team in FFD/ILRI aimed to exploit new opportunities brought by the new advancement in the science of molecular technologies to enhance effective utilization and efficient conservation of these neglected tropical forage species. We have been using genotyping by sequencing (GBS), the oxford nanopore, PacBio, and Hi-C sequencing. For the GBS markers, we are mainly use the DArTseq platform of the Australian diversity array technology, and we have the copy of this platform at ILRI, SEQART Africa. We have generated over 2 million high density genome wide markers by using this technology on few of the tropical forage species. We also conduct field evaluations in replicated field trials and identify best bets for their agronomic performance, water use efficiency, and feed quality traits. By using the high-density genome-wide markers and phenotype data, we have reported the existence of within species genetic diversity and identified QTL/genomic regions associated with the important agronomic and feed quality traits, which are important information in support of the development of forage improved varieties. Last year, we reported a high-quality chromosome-scale assembly of the lablab genome on Nature communication. Other activities, such as identifying potential duplicates (redundant genotypes, by identifying duplicates and removing redundant genotypes, the genebank can reduce cost of maintenance or conservation) and unique genotypes, and subsets or core collection (a few accessions that represent/capture the genetic diversity of the whole collection and easily distribute to users/requesters by the genebank), are important to enhance the utilization and efficient conservation of tropical forage genetic resources maintained in the ILRI forage genebank. Recently, in collaboration with the CIAT forage breeding team and NARS, we have started working on genomic prediction to accelerate selection of best-performing forage accessions from the genebank collections, as well as for accelerated selection of promising progenies to advanced stage in a breeding program using markers-assisted- selection (MAS) and genomic selection (GS). About Meki Shehabu Muktar Meki Shehabu Muktar is a scientist on forage diversity at the International Livestock Research Institute (ILRI). His research focuses on forage genetic diversity, marker-assisted selection, QTL mapping, candidate genes, and genome-wide association mapping in order to identify genes and QTLs controlling agronomic and forage nutritional quality traits. He has been working on genetic diversity, subsetting, and core collection identification, and field evaluation to enhance the utilization and efficient conservation of tropical forage species in the ILRI forage genebank. He is also working on genomic prediction to accelerate selection of best-performing forage accessions from the genebank collections, as well as for accelerated selection of promising progenies to advanced stage in a forage breeding program using markers-assisted-selection (MAS) and genomic selection (GS). His PhD was on plant molecular genetics of tetraploid potato at the MaxPlanck Institute for plant breeding research, Cologne, Germany. Prior to joining ILRI in 2017, he worked at Obihiro University of Agriculture and Veterinary Medicine in Japan as a post-doctoral researcher studying potato genetics at the Potato Germplasm Enhancement Laboratory. January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 6 3. Genomics and genetics research on forage species at the Earlham Institute (EI)-Dr. Jose De Vega, Earlham Institute Recent advances in genomic selection (GS), long-read sequencing (HiFi and ONT reads), and low- coverage whole genome sequencing (lcWGS) are pivotal for enhancing breeding efficiency in Urochloa forage species. Genomic selection minimises the need for extensive field trials by allowing breeding indices and phenotypic values to be predicted through genotypic data derived from sufficiently dense marker maps. However, questions persist about what defines a suitably dense marker set and how this, along with population sizes, impacts prediction accuracy. The cost-effectiveness of lcWGS, especially at approximately 1X coverage, makes it a viable alternative to genome reduction strategies (GBS, RADseq, RNAseq) or SNP arrays, providing accurate marker imputation based on progenitors’ genotype data. In recurrent selection schemes, like the one employed by CIAT’s interspecific Urochloa programme, genetic gains are enhanced by recombining closed gene pools through cycles of phenotypic selection. A strategy was implemented that involved genotyping the programme's founder gene pool at high coverage (20X) and the progeny from the latest cycle (Br19 from 2019) at low coverage (under 1X) to achieve cost-effective genotyping. Bioinformatics pipelines were utilized to impute markers in offspring populations, taking advantage of statistical accuracy even with low sequencing depths. This marker panel was subsequently employed for further analysis at CIAT, ILRI and EI (results presented). Additionally, the advantages of heterozygous allotetraploid genomes, such as in U. decumbens and U. humidicola, were discussed for genetic analysis and breeding. The challenges of assembling complex genomes effectively can be tackled, irrespective of ploidy and heterozygosity levels, by utilising advanced long-read technologies with enhanced error correction profiles, e.g. HiFi reads. It was proposed to test the new ONT kits referred to as Q20+ or v14. This combination of accurate long reads, with Hi-C scaffolding and pruning techniques to eliminate Hi-C connections between homologous or allelic sequences (like Allhic and Haphic), helps to clarify heterozygosity and differentiate homologous sequences on chromosomes. The practical uses of these advancements include mapping traits like root morphology and identifying QTLs related to stress tolerance (e.g., in acidic soils) for marker-assisted selection or trait introgression. The integration of genomic selection with improved genome assemblies enhances the efficiency of hybrid breeding programmes and speeds up the genetic enhancement of tropical forages. Ultimately, these tools tackle the challenges of polyploidy and reproductive complexity, allowing for the effective utilisation of genetic diversity in Urochloa improvement. Jose De Vega is primarily funded by the Biotechnology and Biological Sciences Research Council (BBSRC), part of UK Research and Innovation (UKRI), via Earlham Institute’s Strategic Programme Grant “Decoding Biodiversity” (BBX011089/1), and its constituent work package BBS/E/ER/230002B (Decode WP2 Genome Enabled Analysis of Diversity to Identify Gene Function, Biosynthetic Pathways, and Variation in Agri/Aquacultural Traits). About Jose De Vega Jose De Vega leads a research program in agricultural biodiversity and genomics at the Earlham Institute (EI), focusing on genomics, bioinformatics, and breeding. He applies Agri-tech research to address global food needs and prioritizes collaboration with researchers and stakeholders, supporting open-source principles. He joined EI in 2013 as a postdoctoral researcher on the red clover genome, and since 2015, has managed several crop genomic projects. His group is involved in two core grants from Research Councils UK, researching genome evolution and crop breeding. They study how selection and adaptation influence crop genomes and diversity, focusing on hybrids and repeatedly domesticated species. Their aim is to expedite the development of improved crop varieties through enhanced efficiency in crop January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 7 improvement. They utilise genomic computational analyses to identify superior cultivars with diverse traits and key agronomic markers, leveraging our expertise in large datasets and association mapping. They collaborate with breeders globally to introduce desirable traits into elite genetic lines, employing marker- assisted (MAS) and genomic selection (GS) in their programmes. 4. Developing genomic tools for key underutilized forage species in Africa – Dr. Abel Teshome Gari, ILRI, Ethiopia The challenges to achieving food, nutritional and economic security remain greatest in Africa today, where livestock production remains a mainstay of agricultural livelihoods, demand for livestock-derived foods is rising, and regular consumption of milk and meat is critical to improving the nutritional status of the poor. An inability of Africa’s small-scale livestock keepers to feed their ruminant animals adequately throughout the year is a major barrier to livestock development. To this end, work is under way to redesign Africa’s traditional planted forages which are expected to play a central role in sustainably nourishing the continents ruminant livestock that are feeding, and nourishing, Africa’s growing human populations. The current research presented focuses on developing genomic resources and modern breeding technologies for the tropical forage species such as Napier grass (or elephant grass), one of Africa’s most important traditional forage grasses. Work is undergoing on improving several of Africa’s neglected and underutilized indigenous food-forage species such as lablab, oats and tef. Our research provides valuable genomic resources for tropical forage research and breeding, facilitating the development of improved cultivars with desirable agronomic and nutritional properties. About Abel Teshome Gari Abel Teshome Gari is currently a scientist at Livestock Genetics, Nutrition and Feed Resources Program in the International Livestock Research Instiute (ILRI). Abel’s current research focuses on developing genomics tools for a crucial tropical forage species in Sub-Sahara Africa (SSA). These projects aim to bring sustainable intensification of the smallholder livestock sector activities in SSA, by generating information and tools that can be used to rapidly select productive genotypes with resilience against drought and disease pressure. He is also a research fellow at John Innes Centre and at Earlham Institute. Abel previously completed his PhD study at Swedish University of Agricultural Sciences (SLU), Sweden. Afterwards, Abel conducted his postdoctoral study under Marie Curie-CAROLINE fellowship program at Teagasc, Ireland, focusing on developing marker assay for self-incompatibility genes in perennial ryegrass (Lolium perenne) at Dr Susanne Barth’s lab. 5. Genetic Diversity and Genome-Wide Association Studies for the Important Agronomic Traits of Oat (Avena sativa L.). Ms. Lidiya Ashenafi, ILRI, Ethiopia Oat (Avena sativa) is a significant annual crop that is cultivated globally for a broad spectrum of applications, primarily for livestock feed and human consumption. Despite its multifaceted uses, global oat production has experienced a decline in recent years due to various factors, including the impact of January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 8 biotic and abiotic stresses induced by rapid climate changes. To mitigate production losses caused by these stresses, it is essential to develop oat varieties that exhibit tolerance to these challenges and resilience to evolving climate conditions consequently, genetic and genomic studies play a crucial role in the development of such resilient varieties. Genetic diversity studies are crucial for genetic both conservation and developing new cultivars with superior traits, such as increased yield, disease resistance, and abiotic stress tolerance, as they provide valuable information to help breeders identify and select the most promising varieties for further development. Genome-wide association study (GWAS) is a popular approach for dissecting the genetic basis of complex traits (such as yield, nutritional quality, and stress tolerance) and to pinpoint the causal loci underlying these complex traits. The primary objective of this study is to assess genetic diversity through SNP markers and pinpoint quantitative trait loci (QTL) associated with agronomically important traits in an association panel of 169 oat genotypes through GWAS. The association panel was phenotyped for a range of traits, including phenology, yield-related, and seed quality traits across three locations over a period of two years. Additionally, the panel was genotyped using Genotyping by Sequencing (GBS). Population structure, Principal Component Analysis (PCA), and hierarchical clustering were performed using 1,823 robust SNPs, employing STRUCTURE, R software, and MEGA software, respectively. Population structure and PCA detected two sub-groups with greater degrees of admixture A GWAS analysis was performed based on 18176 high-quality SNPs. Marker-trait associations were computed in R software using a genome association and prediction integrated tool (GAPIT) package and identified 31 marker trait associations (MTAs) that were significantly (false discovery rate, FDR, <0.05) associated with different agronomic traits (cover, total fresh weight, seed length, number of tillers per plant, number of leaves for plant, seed length and stem thickness). The SNP markers identified in this study provide insights into the genetic control of desirable agronomic traits and could be used in genomics-assisted breeding to develop highly productive and climate-resilient oat varieties. About Lidiya Ashenafi Lidiya Ashenafi is an MSc graduate fellow at the International Livestock Research Institute (ILRI), currently conducting her MSc thesis titled “Genetic Diversity and Genome-Wide Association of Important Agronomic Traits in Oats.” She holds a BSc in Biotechnology and previously worked as a Junior and Assistant Researcher at the National Agricultural Biotechnology Research Center (NABRC) from 2019 to 2021. Her current research focuses on understanding genetic diversity and conducting genome-wide association studies (GWAS) to identify quantitative trait loci (QTLs) associated with key agronomic traits in oats. This work is crucial for enhancing breeding programs aimed at developing high-yielding, climate- resilient oat varieties through the incorporation of diverse genetic resources. 6. Use of Genomic Predictions at the Global Barley Breeding Program – Dr. Miguel Sanchez- Garcia The new extreme climatic events due to Climate Change are affecting large portions of the 20 Mha devoted to barley in the Developing World, requiring novel approaches to develop better adapted varieties faster.Conventional breeding programs must choose between increasing testing locations or entries number, especially at early testing stages, due to budget and logistics. Implications are greater for international breeding programs targeting a wide range of environmental and agro-ecological conditions. To overcome this, the Global Barley Breeding Program of the CGIAR implements since 2021 a genomic-assisted sparse multi-location approach, an experimental design that allows breeders to test January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 9 thousands of entries under multi-location trials having only a fraction of the genotypes present at each environment. Low density genotyping is then used to predict traits of interest of non-planted lines across 8 locations in 4 countries representing relevant growing areas. For instance, the high prediction accuracy for yield of these genomic models (r=0.20-0.55), can increase selection accuracy up to 800% as compared to conventional approaches. About Miguel Sanchez-Garcia Dr. Miguel Sanchez-Garcia is the Leader of Global Barley Breeding Program, International Center for Agricultural Research in the Dry Areas (ICARDA) since 2020. Dr. Sanchez-Garcia has more than 15 years’ experience in cereal breeding and research and ten years’ experience in breeding for developing countries. Experienced in classical and molecular approaches to plant breeding, including biotic and abiotic stress tolerance and end-use product quality assessment. Strong innovation spirit to implement new breeding strategies. 7. Genomic prediction in multi-environment and genotype-environment interactions models– Dr. Claudia Samantha Perea, Alliance Bioversity-CIAT This presentation discusses the importance of analyzing genotype-by-environment (G×E) interactions in multi-environment trials (MET) to enable the implementation of genomic prediction (GP) in the Urochloa breeding program. This analysis is a valuable tool for identifying and selecting stable, high-performing genotypes across diverse and contrasting environments while quantifying the environmental effects on trait expression. By incorporating G×E analysis, breeders can optimize resource allocation and make informed decisions to address the challenges of breeding in complex and variable environments. The main statistical tools developed for this purpose were outlined and the cross-validation scenarios for testing these models were described. Additionally, it was presented the first application of this GP approach in a MET of the BR19 Urochloa interspecific population, sequenced using the Whole Genome Sequencing Skim-seq method and characterized for biomass-related traits in three environments. The presentation included the correlation between traits and environments, as well as insights into environmental effects, genotypic effects, and biplots of G×E interactions derived from Singular Value Decomposition (SVD) of the Z-matrix. This method provides a visual representation of the extent and direction of interactions in each environment. GP was implemented using mixed Bayesian models with a block diagonal structure to account for G×E variation across three locations. The analysis included the decomposition of phenotypic variance and the estimation of narrow-sense heritability for each trait, which ranged from 0.15 to 0.46, with plant height showing the highest heritability and area the lowest. The models were validated using two types of 10- fold cross-validation scenarios: one simulating complete field designs (CV1) and the other simulating incomplete designs (CV2). The incomplete design (CV2) demonstrated superior predictive accuracies, achieving 0.68, 0.88, and 0.50 for area, biomass, and height, respectively. These results underscore the potential of this approach for efficient sparse testing design applications in the Urochloa breeding program. Finally, the methods used to design the first low-density SNP panel for quality assurance and control (QA/QC) in the Urochloa interspecific breeding program were introduced. This panel includes 159 KASP sparse markers selected based on minor allele frequency (MAF), quality metrics, and linkage January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 10 disequilibrium (LD). It also incorporated nine trait-related SNPs associated with reproductive mode in forages. This SNP panel is currently undergoing validation, marking a significant step forward in improving the efficiency of the breeding program. About Claudia Samantha Perea Claudia Samantha Perea is the bioinformatician for the Forages Breeding Program at the International Center for Tropical Agriculture (CIAT) in Cali, Colombia since 2022. With a background in biology and a Master's degree in Biological Sciences, she specializes in applying advanced bioinformatics approaches to agricultural research. Her current research focuses on implementing genomic tools to collaborate with the breeding team to enhance breeding efficiency and genetic gain in forage grasses. This includes developing innovative genotyping resources, utilizing genomic data to characterize breeding populations, and optimizing breeding strategies by testing and integrating genotype-by-environment interactions into genomic prediction models to collaborate in the novel implementation of Genomic Selection in the program. As a member of CIAT's Genocrop group for Lever 6 “Crops for Nutrition and Health”, she actively contributes to collaborative efforts that integrate cutting-edge bioinformatics solutions with traditional breeding approaches to accelerate crop improvement programs, particularly focusing on tropical forage species. 8. Development of a mid-density SNP panel for genomic prediction in interspecific hybrids selection in Urochloa spp. – Dr. James Brett, – Earlham Institute Genomic selection (GS) is a powerful tool for improving quantitative traits in Brachiaria species, an allotetraploid forage grass of agricultural importance. This study evaluated the performance of a mid- density marker panel in genomic prediction. Marker data were derived from low-coverage skim sequencing (1X WGS) of the BR19 breeding panel and evaluated against phenotypic traits such as height, area, and biomass across 365 lines. Three marker selection models, top GWAS, balanced GWAS and random —were used to compare prediction accuracies while varying marker densities. The results demonstrated that random marker selection slightly outperformed GWAS-based marker selection in terms of predictive power and error minimisation, though not to a significant degree. Reducing marker density from 5000 to 750 random markers had minimal impact on prediction accuracy, with Pearson correlation decreasing by only ~0.04 and RMSE remaining nearly unchanged across all traits, suggesting that 750 markers can effectively approximate the predictive power of larger marker sets. Below this threshold (e.g., 100 markers), prediction accuracy declined significantly. In conclusion, a mid-density panel of 750 randomly distributed SNPs provides an optimal balance between cost efficiency and prediction accuracy in GS applications for Brachiaria. Further reductions in marker density compromise accuracy, highlighting the need for careful marker selection strategies to maintain reliable genomic predictions. James Brett is primarily funded by the Biotechnology and Biological Sciences Research Council (BBSRC), part of UK Research and Innovation (UKRI), via Earlham Institute’s Strategic Programme Grant “Decoding Biodiversity” (BBX011089/1), and its constituent work package BBS/E/ER/230002B (Decode WP2 Genome Enabled Analysis of Diversity to Identify Gene Function, Biosynthetic Pathways, and Variation in Agri/Aquacultural Traits). January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 11 About James Brett James Brett joined the Earlham Institute in Norwich as a postdoctoral researcher in the De Vega group in 2024. He specializes in crop genomics, with a focus on enhancing breeding and crop improvement by targeting agronomically and nutritionally important traits, such as protein content and dietary fiber. His goal is to collaborate with breeders to effectively apply agricultural research to their breeding programs. As part of the EU Horizon-funded Legume Generation project, James acts as data manager for the consortium, facilitating bioinformatic analyses across the project. Prior to this, his PhD research focused on quantitative trait loci (QTL) identification for the nutritional trait dietary fiber in wheat. This work, as part of the EU FIBRAXFUN project, involved genetic and bioinformatic analyses to identify causal genes and enable wheat breeders to use marker-assisted selection (MAS) to improve wheat nutrition. 9. Genome-Wide Association Study and Genomic Prediction of Forage Biomass-Related Traits in Urochloa spp. – Dr. Haileslassie Gebremeskel, ILRI, Ethiopia The genus Urochloa (syn. Brachiaria), a pivotal contributor to tropical and subtropical livestock nutrition, presents formidable breeding complexities owing to its polyploid nature, pronounced heterozygosity, and predominantly apomictic mode of reproduction. This study employs an integrative approach combining Genome-wide Association Studies (GWAS) and Genomic Prediction (GP) to decipher the genetic architecture underpinning biomass-related traits in 373 Urochloa interspecific hybrids. Utilizing a dataset comprising 32.4K high-quality SNP markers, GWAS uncovered significant loci linked to area coverage, biomass yield, and plant height (PH). Remarkably, the SNP marker M30695 on Chr08 exhibited significant associations with both area coverage and biomass yield, indicative of pleiotropic effects. Additionally, markers M27970, M27973, and M28106 on Chr05 were uniquely associated with biomass, while M04152 on Chr15 demonstrated strong relevance to PH, underscoring its potential utility in breeding programs. Parallel GP analyses employed an array of models, including BLUP (via the rrBLUP R package), Bayesian methods (BGLR R package), and advanced machine learning and deep learning algorithms (utilizing Keras and scikit-learn in Python). Among these, machine learning and deep learning frameworks, such as CNN, RKHS, Ridge Regression, PLS Regression, ElasticNetCV, SVR-linear, SVR-poly, and Linear Regression, emerged as the most robust predictors for area coverage, biomass yield, and PH. Furthermore, the study evaluated the influence of training population (TP) sizes and SNP densities on prediction accuracy for the four traits, revealing that the top5000 SNP markers achieved the highest prediction accuracy, closely followed by the significant loci identified through GWAS. This research provides pivotal insights into the genetic underpinnings of key agronomic traits in Urochloa interspecific hybrids. It highlights the synergistic potential of integrating GWAS and GP to expedite breeding initiatives. The identification of establish SNP markers and corresponding prediction accuracies January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 12 establishes a solid foundation for marker-assisted selection (MAS), genomic selection, and the development of superior Urochloa hybrids tailored for enhanced agronomic performance. About Haileslassie Gebremeskel Haileslassie Gebremeskel is an accomplished molecular breeder and geneticist with expertise in vegetable breeding and genetics. Holding a Ph.D. in Vegetable Sciences with a focus on molecular breeding and genetics from the Chinese Academy of Agricultural Sciences (CAAS), he has established himself as a senior researcher at the Ethiopian Institute of Agricultural Research (EIAR). Currently, he serves as a research fellow at the International Livestock Research Institute (ILRI), where he leverages genomic and bioinformatics tools to advance the improvement of forage grasses in Africa. Haileslassie’s research delves into genetic diversity assessment, genome-wide association studies (GWAS), and the development of genomic selection models for Urochloa interspecific breeding. His efforts aim to enhance the accuracy and efficiency of selection in Urochloa hybrids. Notably, his work targets genetic diversity analysis, GWAS, and GS. Presently, Haileslassie is identifying markers significantly associated with area coverage, biomass yield, and plant height, genomic prediction models that improves accuracy, reduces breeding cycles, lowers costs by minimizing field trials, enables large-scale screening, and enhances selection intensity of Urochloa interspecific hybrids. This project seeks to accelerate genetic progress in breeding programs by reducing phenotypic evaluation time and increasing selection intensity, employing marker-assisted or genomic selection methodologies at early breeding stages. 10. Genome-wide association study and genomic prediction for important traits in guinea grass (Megathyrsus maximus) – Dr. Kefyalew Negisho Bayissa, ILRI, Ethiopia Guinea grass (Megathyrsus maximus) is an important forage grass that plays a significant role in pasture- based livestock production systems to feed animals, especially in tropical and subtropical regions. It is a leafy perennial tropical C4 and climate-resilient crop with biological nitrification inhibition and CO2 fixation potential. However, in traditional breeding schemes, accurate measurement of productivity traits limits its genetic gain. This study aimed to identify genetic variants associated with important traits and explore genomic prediction for molecular breeding in Megathyrsus maximus. Phenotype data converted to BLUP value for Plant Height (PH), Biomass (BM), Area Cover (AC), and Normalized Difference Vegetation Index (NDVI) traits from three locations (LIanos, Palmira, and Quili) were used in combination with 8,133 filtered SNP markers distributed across the genome. We employed GAPIT3 for GWAS analysis, BLUP (gBLUP, cBLUP, and rrBLUP) and Bayesian (BL, BRR, and BayesB) models to optimize prediction accuracy for genomic prediction. Whole markers and markers with P<0.05 were tested using prediction models. The study panel was clustered into three populations with PC1 and PC2 explaining 34.16% and 11.36% of the genetic variation, respectively. A total of 49 significant MTAs were detected at 5% FDR, associated with the BLUP values of traits across the three locations and the combined analysis, of which January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 13 7, 6, 5, and 4 were associated with PH, BM, AC, and NDVI combined analysis, respectively. The identified MTAs were clustered into 33 QTLs using LD critical r2 = 0.2. BLASTn analysis depicted biological functions and processes for some of the significant QTLs linked to genomic regions while others might be new findings. Regardless of the training populations, models, and investigated traits, markers with p-value < 0.05 increased the prediction accuracy compared to the whole markers. Mean separation revealed that gBLUP outperformed other models and was used to estimate genomic estimated breeding values (GEBVs). The results provide valuable information for improving Megathyrsus maximus through molecular breeding. About Kefyalew Negisho Bayissa Kefyalew Negisho Bayissa obtained his PhD in agriculture from Martin Luther University of Halle- Wittenberg, Germany and his MSc in International Horticulture majoring in plant nutrition at Leibniz University of Hannover, Germany, where he gained skills and experiences in plant nutrition on horticultural crops, plant breeding and genetics, bioinformatics, genetic engineering, and R software. He obtained his BSC in plant sciences from Haromaya University. During his PhD study, he coordinated a research project for the Ethiopian Institute of Agricultural Research (EIAR) on “Genome-Wide Association Studies to Improve Drought Stress Tolerance in Ethiopian Durum Wheat (T. durum) and Barley (H. vulgare) Landraces”. Dr. Kefyalew authored and co-authored six articles from the PhD project in peer-reviewed journals on durum wheat and barley. He has excellent experience in executing research projects as a PI and Co-PI. He won prestigious awards, including the German Academic Exchange Service (DAAD) fellowship and the Norman E. Borlaug Fellowship International from Michigan State University. He served as a wheat breeder at Kulumsa Agricultural Research Center, as a maize breeder at Bako National Maize Program, and on plant pathology at Ambo Agricultural Research Center, EIAR. Dr. Kefyalew was a national coordinator of the National Plant Biotechnology and Crop Protection programs in EIAR for three years, deputy director of the Agricultural Biotechnology Research Sector, and acting as center director for the National Agricultural Biotechnology Research. He worked as a consultant with the International Food Policy Research Institute (IFPRI) and the International Maize and Wheat Improvement Center (CIMMYT) on maize and wheat for project development for baseline assessment of tracking crop varieties through DNA fingerprinting. He had made a dialog with policymakers on biosafety for the safe use of biotech technologies and he organized training and workshops to create awareness for the sustainable use of biotech research outputs. Dr. Kefyalew is serving at large as a mentor and supervisor in his discipline for several postgraduate students. Currently, he has been working as a Research Fellow at ILRI Feed and Forage Development (FFD) research program since October 2023, in Addis Ababa, Ethiopia. He focuses on phenotypic and genotypic data analysis with R, Linux, and Phyton using the ILRI HPC. 11. Genomic Selection in Napier grass. – Dr. Zewdinesh Damtew Zigene, ILRI, Ethiopia Napier grass (Pennisetum purpureum) is a vital forage crop in tropical regions due to its high biomass yield and adaptability. To improve its breeding efficiency, genomic selection (GS) is a promising tool that uses dense genomic marker data to predict complex traits. This study was designed to develop a genomic January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 14 tool to support the application of next generation genomics-based selection and breeding strategies in Napier grass, and to evaluate the efficiency and accuracy of three genomic selection models. This study involved analysis of phenotypic and genotypic data from 84 Napier grass to evaluate the performance of GS models and their potential in breeding programs. The dataset consisted of 600,063 SNPs markers obtained through whole genome sequence, and nine field phenotype data. Quality control steps for genotypic data, such as filtering SNPs with minor allele frequency (MAF < 0.05) and missing data >20%, resulted in a refined dataset. Ridge Regression Best Linear Unbiased Prediction (rrBLUP), Genomic Best Linear Unbiased Prediction (gBLUP) and Conventional Best Linear Unbiased Prediction (cBLUP) models were used for analysis. Cross-validation (5-fold) was employed to evaluate model accuracy, with prediction performance assessed using the correlation between observed and predicted phenotypes. After quality filtering, 138,767 SNPs were retained for analysis. The rrBLUP and gBLUP models demonstrated robust predictive accuracy, with rrBLUP achieving slightly higher accuracy. The analysis indicated their potential for marker-assisted breeding. The study demonstrated the effectiveness of genomic selection models in predicting complex traits in Napier grass using high-density SNP data. This approach provides valuable insights for improving breeding efficiency and developing superior Napier grass varieties. Future work should integrate phenotypic data and expand to larger populations to validate and refine predictions. About Zewdinesh Damtew Zigene Zewdinesh Damtew is an agricultural researcher specializing in aromatic and medicinal plants breeding. She graduated with PhD degree in plant science focusing on molecular breeding from Hawassa University and works as a senior researcher in plant biotechnology program at the Ethiopian Institute of Agricultural Research. Throughout her career, she has developed a deep passion for crop management, field phenotyping and molecular genetics as well as acquired experience in genomic DNA extraction, PCR, genomic data analysis, in vitro techniques, and scientific paper writing. Currently she is working as a research fellow in the Livestock genetics, Nutrition and Feed resource program at the International Livestock Research Institute. Her work focuses on genomic selection and genomic prediction of Napier grass. Zewdinesh is contributing to genomic selection of Napier grass by optimizing different genomic selection models. 12. Genomic Prediction of Forage Biomass-Related Traits in Oats (Avena Sativa) and Lablab (Lablab purpureus). – Mr. Hailu Lire Wachamo, ILRI, Ethiopia Livestock production is crucial for global livelihoods and food security, supporting 1.3 billion people1. Sub-Saharan Africa accounts for 14% of the world's livestock, but its production hindered by various challenges2. Tropical forages, including grasses, legumes, and multipurpose trees, are common forages to improve productivity and income3. The lack of high-quality feed limits the efficiency of smallholder crop-livestock operations 4. Genomic selection uses genomic data to predict traits and select individuals with the desired characteristics5. Commonly used in plant breeding schemes, to improve crop yields, January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 15 disease resistance, and other important traits through (1) genotyping (measuring genetic variation), (2) phenotyping (measuring traits), (3) genome-wide association studies (GWAS) (identifying genetic variants linked to traits), and (4) predictive modeling (using statistical models to predict phenotype from genomic data)6. Recent studies have found QTLs linked to yield, forage quality, and stress resilience in crops like Napier grass7. Genome-Wide Association Studies (GWAS) were used to scan forage crop genomes for genetic variations related to traits like quality, yield, and disease resistance8. Genomic selection also plays a role in breeding programs, predicting breeding values to select individuals with desirable traits, enhancing yield, quality, and resistance 5. Lablab (Lablab purpureus) and oat (Avena sativa), key livestock feed crops, face challenges from abiotic stresses like frost, salinity, and waterlogging, and biotic stresses like pests and diseases, all worsened by climate change and advanced tools like genomic selection were important to improve their production. The main aim of this fellowship was to develop genomic tools for forages (Lablab and oats) to enhance genetic gains by estimating their genetic/breeding value using proper genomic selection tools. Seven genomic selection tools are developed and applied; different statistical models were evaluated for accuracy and assessed using correlation and R² for prediction quality, MSE or RMSE for error magnitude, and accuracy compared to heritability for selection efficiency. The rrBLUP, RF, BL, BA, BC, BB, and SVR are among the models that are found best for predicting breeding values (GEBVs) in forage species with accurate phenotypic and genotypic data. Hailu Lire Wachamo Hailu Lire Wachamo is an early-career researcher specializing in plant molecular breeding programs. He holds a BSc in Plant Science, an MSc in Plant Biotechnology, and an MSc in Horticultural Science from Wolaita Sodo, Hawassa, and Jimma University, respectively. He works as a Research Fellow at the Research & Innovation, Livestock Genetics, Nutrition, and Feed Resources program at the International Livestock Research Institute. Previously, he was a researcher in the plant biotechnology research program at the Ethiopian Institute of Agricultural Research (EIAR) under the Agricultural Biotechnology Directorate. His current research focuses on accelerating breeding progress through applying genomic selection tools in legumes like Lablab, and forage grasses. Currently, he applies genomic selection tools to improve forages, such as Lablab purpureus and Avena sativa L., aiming to enhance genetic gains and improve selection efficiency. January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 16 13. Progress on QTL mapping for spittlebug resistance in interspecific F1 biparental families of Urochloa spp. based on the RADseq sequence data. Dr. Paula Andrea Espitia, Alliance Bioversity- CIAT Spittlebugs (Hemiptera: Cercopidae) are the major production constraint for grass cash crops in the tropical and subtropical regions of the Americas, causing economic losses estimated at US$840–2,100 million annually across the sugarcane and forage grass value chains. Resistance to spittlebugs remains a key breeding objective in Urochloa programs at CIAT. Hybrids and parental lines are screened for antibiosis and tolerance using no-choice tests under screenhouse conditions, achieving high genetic gains. However, these trials are resource-intensive, requiring skilled labour to maintain spittlebug colonies and synchronize plants and insects, and are limited to around 150 genotypes per experiment with four replicates, creating bottlenecks for scaling or early-stage integration. Molecular techniques like marker-assisted selection and genomic selection have the potential to improve the selection process for Urochloa hybrids and parental lines. Integrating these techniques into the breeding pipeline could improve genetic gains by reducing the time needed for resistance trials and enabling earlier-stage selection. The main objective for this project is to construct genetic maps of an Urochloa interspecific mapping population and identify QTL associated with resistance to Aeneolamia varia nymphs. A major challenge, is that QTL discovered in single biparental populations, derived from highly heterozygous outbred individuals can lose their predictive ability when applied to the wider breeding populations. To address this, a pseudo–Nested Association Mapping (NAM) F1 outcross population was developed by crossing four resistant sexual genotypes from the 10th recurrent selection cycle with Urochloa decumbens, a susceptible apomictic tester. This resulted in four families comprising over 300 genotypes, evaluated for antibiosis and tolerance using the screening protocol to estimate the plant damage (quantified by image analysis) and insect survival, and analyzed using ASReml to estimate BLUEs and BLUPs through spatial mixed models, and genotypes were sequenced twice using RAD technology to increase read depth. Aligning to a Urochloa ruziziensis haploid reference produced three times more variants and higher read depth compared to the Urochloa decumbens tetraploid reference, due to the stacking of reads aligning to a single location. Variant calling was performed using a tetraploid setup applied to the diploid-aligned reads, allowing the estimation of allele dosages probabilities with the updog R package. Among the models tested, the normal prior distribution reduced overdispersion and allele bias, obtaining more accurate results. However, several markers showed deviations in segregation ratios, which is expected given the allotetraploid nature of the population. Polyploid-specific tools like Mappoly and polymapR were used for linkage mapping, but issues like segregation distortion and a shortage of simplex marker prevented progress for following the outcrossing population approach. Consequently, a diploid approach was used for advancing the genetic mapping in Onemap R package, which simplified the analysis. Next steps involve refining genetic maps using different marker ordering algorithms and associating phenotypic data of plant damage and insect survival with the updated maps for the QTL discovery. About Paula Andrea Espitia Paula Espitia-Buitrago is an agronomist and early-career researcher specializing in Urochloa breeding. She holds an M.Sc. in Agricultural Sciences with a focus on Plant Breeding from Universidad Nacional de Colombia and works as an associate researcher in the Tropical Forages Program at the Alliance of Bioversity International and CIAT. Paula’s work focuses on developing and refining phenotyping January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 17 methodologies for Urochloa interspecific breeding trials, leveraging image-based techniques to improve the accuracy and efficiency of high-throughput protocols. Her research emphasizes host-plant resistance to key pests, such as spittlebugs and spider mites. Currently, Paula is contributing to a QTL mapping project aimed at identifying genomic regions associated with Aeneolamia varia nymph resistance. This research seeks to accelerate genetic gains in the breeding program by reducing evaluation time and increasing selection intensity through marker-assisted or genomic selection at earlier stages of the breeding scheme. January 25 | Building capacity and resources for genomic prediction in forage grass and legume breeding 18 The CGIAR Research Initiative on Accelerated Breeding aims to develop better-performing, farmer- preferred crop varieties and to decrease the average age of varieties in farmers’ fields, providing real-time adaptation to climate change, evolving markets and production systems. It forms part of CGIAR’s new Research Portfolio, delivering science and innovation to transform food, land, and water systems in a climate crisis. This report was produced as part of the CGIAR initiative on Accelerated Breeding which is supported by contributors to the CGIAR Trust Fund. This document is licensed for use under the Creative Commons Attribution 4.0 International Licence. Month Year | | | https://www.cgiar.org/funders/ https://www.cgiar.org/funders/ https://twitter.com/CGIAR?ref_src=twsrc%5Egoogle%7Ctwcamp%5Eserp%7Ctwgr%5Eauthor https://www.facebook.com/onecgiar/ https://www.linkedin.com/company/cgiar https://www.youtube.com/channel/UCYuSEwWKAsoNwg6MJEI-qeA