Balancing gain and diversity: Optimal contribution selection for sustainable genetic improvement in sugarcane Authors: Natalya Vo Van‐Zivkovic, Ben J Hayes, Karen Aitken, Eric Dinglasan, Terry Morgan, Seema Yadav Published: 2026 ABSTRACT AbstractSugarcane (Saccharum spp. hybrids), the source of ∼80% of the worlds sugar, has experienced a plateau in genetic improvement over recent decades. This stagnation is largely due to the crop's complex polyploid genome and long breeding cycles, which limit the efficiency of traditional selection methods. Genomic selection (GS) has been proposed to accelerate breeding progress in sugarcane by enabling faster improvement of key commercial traits such as cane yield, sugar content, and disease resistance. However, especially in clonally propagated outbred crops like sugarcane, intensive GS can reduce genetic diversity. Prioritizing short‐term genetic gain without managing diversity risks inbreeding depression, resulting in reduced vigor and decreased productivity and compromising long‐term gains and potential for adaptability to changing environments. Optimal contribution selection (OCS) offers a strategic solution by optimizing parental contributions to maximize genetic gain while controlling the rate of inbreeding. This review outlines the challenges in sugarcane breeding, including high ploidy, complex genetic architecture, and low trait heritabilities. We propose the first conceptual framework for applying OCS in sugarcane, highlighting its potential to achieve sustainable genetic improvement. Furthermore, we explore extensions to the OCS framework that incorporate non‐additive genetic effects (e.g., dominance and epistasis), allele dosages, and the multi‐species germplasm that underpins modern cultivars. By balancing genetic gain with diversity retention, OCS provides a pathway for long‐term progress in sugarcane breeding. In the context of rising global sugar demand and climate change, this balanced approach supports improved yield, quality, and resilience while maintaining the genetic diversity essential for future gains. Abbreviations best linear unbiased prediction clonal assessment trial commercial cane sugar genomic estimated breeding value genomic selection linkage disequilibrium optimal contribution selection predicted clonal performance specific combining ability single‐nucleotide polymorphism tonnes of cane per hectare INTRODUCTION Sugarcane (Saccharum spp. hybrids) is an economically important crop grown in tropical and subtropical areas, contributing to over 80% of global sugar production (FAOSTAT, 2026). It is also emerging as an important bioenergy source due to its inherently high biomass production capacity, a growing area of interest for sustainable agriculture (Lu et al., 2024). As global populations rise and dietary preferences shift, demand for sugar and bioenergy will continue to increase. Despite its economic importance, the rate of genetic improvement has stagnated in recent years, leading to a critical need for innovative breeding methodology. Although production has increased over the last few decades, this was achieved through improvements in agronomic practices and the successful development of better adapted varieties (Bordonal et al., 2018; Yadav et al., 2020). Whilst global yields have approximately doubled since the 1960s, recent gains are marginal, challenged by the complexity of its polyploid genome, low trait heritabilities, and long‐breeding cycles (FAO, 2025; Voss‐Fels et al., 2021; Wei & Jackson, 2016; Yadav et al., 2020). Increasing the rate of genetic gain for key traits like cane yield, sugar content, and stress resistance is widely recognized as a long‐term, permanent, and sustainable solution to increasing productivity (Yadav et al., 2020). Four major challenges define the current breeding dilemma for sugarcane: an exceptionally complex, highly polyploid genome affected by aneuploidy; low narrow‐sense heritability for major economic traits; long breeding cycles; and a narrowing genetic base within elite germplasm. Collectively, these challenges limit the efficacy of traditional selection methods and hinder development of new, superior cultivars. Furthermore, the complex quantitative inheritance and perennial nature of sugarcane restrict breeders’ ability to achieving substantial gains. Thus, understanding these constraints is crucial for designing breeding strategies capable of sustaining genetic progress in sugarcane. This review will synthesize the current knowledge on sugarcane genetics, breeding challenges and advanced selection methods by: summarizing the challenges in sugarcane breeding and discussing why these features limit conventional breeding; summarizing the challenges in sugarcane breeding and discussing why these features limit conventional breeding; assessing the effectiveness of genomic selection (GS) and the risk of diversity loss under truncation selection; assessing the effectiveness of genomic selection (GS) and the risk of diversity loss under truncation selection; introducing optimal contribution selection (OCS), a method that aims to maximize long‐term genetic gain while controlling inbreeding and evaluate its successful use in other species; introducing optimal contribution selection (OCS), a method that aims to maximize long‐term genetic gain while controlling inbreeding and evaluate its successful use in other species; providing comparisons of OCS with other widespread selection strategies such as phenotypic selection and GS; providing comparisons of OCS with other widespread selection strategies such as phenotypic selection and GS; proposing the first conceptual framework for applying OCS in sugarcane and offer extensions to the framework that address key challenges by accounting for non‐additive genetic effects, allele dosages, and multi‐species germplasms; and proposing the first conceptual framework for applying OCS in sugarcane and offer extensions to the framework that address key challenges by accounting for non‐additive genetic effects, allele dosages, and multi‐species germplasms; and presenting future prospects and research priorities for implementing OCS in sugarcane breeding. presenting future prospects and research priorities for implementing OCS in sugarcane breeding. GENOMIC COMPLEXITIES OF SUGARCANE Modern sugarcane cultivars have complex polyploid genomes Modern sugarcane cultivars are mostly derived from interspecific hybridization between two Saccharum species, Saccharum spontaneum and Saccharum officinarum, followed by continuous backcrossing to either S. officinarum or a hybrid variety (Aitken, 2022). The objective was to combine high sugar content and good milling traits from S. officinarum with the robustness, disease resistance, and ratooning ability of S. spontaneum (Yadav et al., 2020). The resulting modern hybrids are mosaic allo‐autopolyploids consisting of ∼80% S. officinarum, ∼10%–20% S. spontaneum, and ∼10% from recombination between the two, with random chromosomal recombination during meiosis leading to variations in chromosome number between genotypes (Aitken, 2022; D'Hont et al., 1996; Piperidis et al., 2010). Consequently, modern sugarcane varieties have 100–120 chromosomes, with ploidy levels of 8–12 and a genome size of 10GB (Aitken, 2022). This polyploidy complicates the effectiveness of classical genetic breeding methods through allele dosage effects, where differing copy numbers of alleles yield different phenotypes (Zhang et al., 2025). Implications of polyploidy on trait architecture Likely due to its polyploid architecture, many important traits in sugarcane are heavily influenced by non‐additive effects (Hogarth et al., 1981). Cane yield and sucrose content are controlled by numerous genes of small effect and also by epistatic interactions between alleles (Aitken et al., 2008; Ming et al., 2002). Moreover, many traits have low narrow‐sense heritabilities where only a small proportion of additive variance (which is passed on to progeny) contributes to the phenotypic variance (Falconer & Mackay, 1996). For example, using an extended GBLUP model, Yadav et al. (2021) reported a significant contribution of non‐additive effects in sucrose and fiber content, and accounted for almost two‐thirds of the total genetic variation for cane yield. Polyploid crops are influenced by allele dosage effects, where multiple allelic copies affect gene expression and phenotypic outcome through additive and non‐additive effects (De Bem Oliveira et al., 2019; Osborn et al., 2003). In many polyploid crops, such as tetraploid potato and blueberry or hexaploid sweet potato, allele dosage can range from zero to four or six copies of an allele, respectively. Packages like PolyRAD have been developed specifically to analyze these levels of polyploidy (Clark et al., 2019). In contrast, modern sugarcane cultivars possess highly complex and variable polyploid genomes, with chromosome numbers ranging from approximately 100–130 and locus‐specific ploidy levels typically varying from 8 to 12 copies (D'Hont et al., 1996; Piperidis et al., 2010). Consequently, a single allele may exist across a much wider dosage spectrum than in other polyploid species, greatly increasing the number of possible genotypic classes and the complexity of estimating dosage effects. However, modeling allele dosages remains challenging because many genotyping platforms are unable to accurately determine allele copy number at a given locus (Dufresne et al., 2014; Njuguna et al., 2023). For example, a study modeling allele dosage in sweet potato and sugarcane found predictive ability to be low in sugarcane but moderate to high in sweet potato (Batista et al., 2022). However, these improvements were non‐significant when applied to both datasets. The authors attribute this to the high proportion of homozygous and single‐dose markers, which limit model sensitivity to dosage information. This was supported by simulations, where populations with greater heterozygosity yielded higher predictive abilities, whereas populations reflecting the predominantly homozygous genotype distributions of the real datasets produced comparable predictive performance (Batista et al., 2022). Notably, these divergent outcomes emerged even between sweet potato and sugarcane, both high‐order polyploids with overlapping ploidy levels. This underscores the extent to which dosage‐related dynamics can vary between polyploid species and highlights why extrapolating findings from lower‐ploidy crops to sugarcane's much higher and more complex ploidy system is unlikely to be appropriate. Modern sugarcane cultivars are derived from interspecific hybridization between multiple Saccharum species, resulting in highly admixed genomes containing chromosome segments from distinct ancestral lineages. Similar to observations in livestock populations, where mutations affecting quantitative traits may have arisen independently in divergent ancestral groups and consequently alter marker effects across ancestries (Warburton et al., 2023), the effect of a genetic variant in sugarcane may depend on the species from which it originated. For example, a single‐nucleotide polymorphism (SNP) associated with sucrose accumulation might have a positive effect when carried on an S. officinarum haplotype but a neutral or negative effect when present on other Saccharum species haplotype. Such differences may arise because causal mutations, linkage disequilibrium (LD) patterns, and local genomic backgrounds differ between the ancestral species (e.g., Warburton et al., 2021, 2023). Ignoring species‐of‐origin information therefore assumes that marker effects are consistent across ancestries, which may reduce the accuracy of genomic prediction and obscure favorable alleles. Thus, accounting for ancestry‐specific effects could enable breeders to better exploit beneficial contributions from different founder species, such as disease resistance or environmental adaptation from S. spontaneum and high sugar content from S. officinarum. BREEDING CHALLENGES AND LIMITATIONS OF CONVENTIONAL SELECTION Long and resource‐intensive breeding cycles Genetic improvements in sugarcane are constrained by long breeding cycles, which typically span 8–12 years across several propagation and selection stages (Wei & Jackson, 2016; Wei et al., 2022; Yadav et al., 2020). The cycle involves multiple stages, described here in an Australian breeding program but following a globally standard framework: (1) cross pollination between selected parents; (2) planting of thousands of seedlings in early‐stage single row plots (progeny assessment trial); (3) selection of best seedlings (often <5%) and propagating them vegetatively into larger plots for further evaluation (clonal assessment trial [CAT]); (4) assessing clones for yield, disease resistance, and ratoon stability across multiple environments and ratoon crops in advanced‐stage trials with large plots to account for competition effects (final assessment trial); and (5) releasing the highest performing clones as varieties and/or for introduction back into the breeding program (Park et al., 2007; Wei et al., 2022) (Figure 1). Candidates are assessed based on key commercial traits, including commercial cane sugar (CCS), fiber content, tonnes of cane per hectare (TCH), and disease resistance (Wei et al., 2022). Accurately measuring these traits makes trial assessment incredibly laborious and time‐consuming, with many breeding programs being constrained by available land and labor required for propagation and phenotyping (Deomano et al., 2020; Wei et al., 2022). Conventional Australian sugarcane breeding program. Best linear unbiased prediction (BLUP) breeding values are used to select parents for crossing. A large number of progeny is produced and subjected to several intense stages of selection, culminating in the commercial propagation and release of one or two superior cultivars, which are introduced back into the breeding program. FAT, final assessment trial; PAT, progeny assessment trial. This long generation interval limits the rate of genetic gain, as indicated by the breeder's equation, where genetic gain per unit time is inversely proportional to the length of the breeding cycle (Walsh & Lynch, 2018). Moreover, early‐stage selection is impacted by small plot sizes and large competition effects, resulting in low selection accuracy and potential misclassification of superior genotypes (Jackson & McRae, 2001). Consequently, the inherently long breeding cycle contributes significantly to the slow improvement in sugarcane yield. Low trait heritabilities complicate genetic gains in sugarcane Genomic improvement in sugarcane is further challenged by the relatively low narrow‐sense heritabilities of key commercial traits. Estimates of narrow‐sense heritability for TCH are ∼0.3, with CCS and fiber slightly higher at ∼0.5 and ∼0.7, respectively (Yadav et al., 2021). Low heritability reduces the effectiveness of selection based solely on individual performance, as much of the observed variation is not transmissible to the progeny, thereby slowing long‐term genetic gain. In contrast, non‐additive genetic effects, such as dominance and epistasis, play a substantial role in determining individual performance. Research has shown these effects to contribute to the variance of key traits in sugarcane such as CCS and TCH (Yadav et al., 2021). These effects arise from interactions among alleles within and between loci and are expressed as specific combining ability (SCA), whereby particular parental combinations produce progeny that perform better or worse than expected based on the average effects of their parents alone (Wei & Jackson, 2016). Consequently, research is increasingly focused on prediction models that integrate both additive and non‐additive effects (Shahi et al., 2025; Voss‐Fels et al., 2021; Yadav et al., 2023, 2021). Exploiting these effects is incredibly beneficial in clonally propagated crops like sugarcane, where non‐additive influences can be preserved through clonal propagation. For example, once a superior clone is identified, it can be clonally propagated to preserve both its additive and non‐additive advantages, allowing its high performance to be preserved. However, despite their high performance, clones selected on total genetic value (additive + non‐additive) often make poor parents because non‐additive effects do not accumulate across generations. Because the primary goal of OCS in breeding programs is to improve the additive genetic gain for long‐term population improvement, accurately modeling these, and balancing them with non‐additive effects, is essential for making appropriate selection decisions. Narrowing genetic base and inbreeding concerns Despite its long history of domestication, there have been only ∼10 recombination events dating back to key ancestors (Aitken, 2022). Consequently, many widely used cultivars grown today are directly descended from older, superior varieties of sugarcane dating as far back as the 1950s (Wei & Jackson, 2016). Inevitably, this heavy use of older, high‐performing varieties has led to similar parentage among a large proportion of clones, culminating as high levels of inbreeding within breeding populations (Hayes et al., 2021). Historical pedigree analyses of elite sugarcane cultivars in Taiwan revealed substantial relatedness within the breeding population, with inbreeding coefficients of 0.281 for ROC10 and 0.156 for ROC9, and pairwise coancestry coefficients reaching 0.203 among six important cultivars (Chang, 1996). These results highlighted the narrow genetic base arising from repeated use of a small number of founding parents. These estimates cannot be assumed to represent contemporary Australian or global sugarcane breeding populations, but they demonstrate that measurable inbreeding and coancestry can accumulate when breeding progress is derived from a restricted set of elite parents. For comparison, an analysis of Australian canola breeding populations assumed an effective population size of approximately 11 and estimated an inbreeding coefficient of 0.21 after five recurrent selection cycles, equivalent to a 21% cumulative loss of genetic variation through random genetic drift (Cowling, 2007). Direct numerical comparisons between canola and sugarcane should be interpreted cautiously because of major differences in mating system, reproductive biology, ploidy level, pedigree structure, and the methods used to estimate inbreeding. Nevertheless, these examples highlight the importance of monitoring genetic diversity within breeding programs. Empirical estimates of realized effective population size, genomic inbreeding, and temporal changes in ancestry remain limited for commercial sugarcane breeding programs. A lack of genetic diversity has been reported as one of the causes for stagnating genetic progress (Lu et al., 2024), and quantifying these parameters in sugarcane is therefore a high research priority to ensure long‐term genetic gain while maintaining sufficient genetic diversity for future breeding. Inbreeding depression arises from the mating of closely related individuals causing a reduction in performance from decreased vigor as a result of fixation of deleterious alleles (Charlesworth & Willis, 2009; Clo & Kolář, 2022; Lin et al., 2017; Merrick et al., 2023). In sugarcane's polyploid genome, multiple copies of deleterious alleles may remain hidden due to masking by dominant alleles, only being exposed after many generations of closely related matings. However, in the long term these deleterious alleles can escape natural selection and rapidly increase in frequency within a population (Clo & Kolář, 2022). Studies have shown that self‐pollination (the most extreme form of inbreeding) negatively impacts agronomically important traits in cross‐pollinated crops, showing a marked decline in quality and yield traits (Ceballos et al., 2015; De Azeredo et al., 2016; Ferreira et al., 2005; Hogarth, 1980; Lin et al., 2017; Nakanishi et al., 2015; Shaw, 1997; Silva & Gonçalves, 2011). Introgression of wild S. spontaneum to increase genetic diversity has been successful in some sugarcane breeding programs (Kimbeng et al., 2021; Yadav et al., 2020). However, extensive LD across the genome, combined with long‐breeding cycles, hinders introgression efforts, making it slow and difficult to separate desirable traits from undesirable ones (Mahadevaiah et al., 2021; Yadav et al., 2020). Consequently, this approach has seen limited adoption by breeders. Phenotypic selection: Limitations in sugarcane In Australia, modern sugarcane breeding programs still largely depend on truncation selection where ranking criterion is based solely on phenotypic records or a pedigree‐based expected breeding value (Wei et al., 2022). Rank‐based selection remains attractive because it is transparent, computationally simple and readily incorporated into crossing programs constrained by flowering, viable crosses, and nursery capacity. Its limitation is that related high‐ranking candidates can be selected together unless relationship and parental contribution are explicitly controlled. Although advantageous for highly heritable traits, truncation selection based on phenotypic values has major limitations in sugarcane: Selection accuracy: As many traits have low narrow‐sense heritability and strong environmental and non‐additive effects, selecting based on raw or adjusted phenotype is unlikely to contribute to long‐term genetic gain. High performance of candidates may instead be attributed to favorable environments and/or specific allele combinations, both of which are not transferred to offspring (Wei & Jackson, 2016). Modeling of various selection methods has shown that truncation selection on phenotypic values can achieve substantial genetic improvement; however, GS frequently delivers greater long‐term genetic gain by increasing selection intensity and reducing the length of the breeding cycle (Gorjanc et al., 2018; Voss‐Fels et al., 2021). Selection accuracy: As many traits have low narrow‐sense heritability and strong environmental and non‐additive effects, selecting based on raw or adjusted phenotype is unlikely to contribute to long‐term genetic gain. High performance of candidates may instead be attributed to favorable environments and/or specific allele combinations, both of which are not transferred to offspring (Wei & Jackson, 2016). Modeling of various selection methods has shown that truncation selection on phenotypic values can achieve substantial genetic improvement; however, GS frequently delivers greater long‐term genetic gain by increasing selection intensity and reducing the length of the breeding cycle (Gorjanc et al., 2018; Voss‐Fels et al., 2021). Cycle length and resource constraints: Multi‐year phenotypic evaluation is required to assess yield, ratoon stability, and genotype‐by‐environment interaction. This prolongs the breeding cycle and limits the number of individuals that can be tested with adequate replication. Cycle length and resource constraints: Multi‐year phenotypic evaluation is required to assess yield, ratoon stability, and genotype‐by‐environment interaction. This prolongs the breeding cycle and limits the number of individuals that can be tested with adequate replication. Genetic bottlenecks: Repeated truncation selection typically advances only the top few percent of clones at each stage, some of which are likely to share alleles (or come from similar families), particularly when selection intensity is high. A study comparing selection methods in sugarcane found that repeated phenotypic truncation selection exhausted total genetic variance within a few breeding cycles (Voss‐Fels et al., 2021). Genetic bottlenecks: Repeated truncation selection typically advances only the top few percent of clones at each stage, some of which are likely to share alleles (or come from similar families), particularly when selection intensity is high. A study comparing selection methods in sugarcane found that repeated phenotypic truncation selection exhausted total genetic variance within a few breeding cycles (Voss‐Fels et al., 2021). Estimated genetic (or adjusted phenotype) values are typically obtained as best linear unbiased predictions (BLUPs), a statistical approach for predicting the genetic merit of individuals from phenotypic observations while accounting for fixed and random effects (Henderson, 1953). Pedigree‐based BLUPs are derived from the analysis of phenotypic and pedigree information and provide a more systematic approach to selection compared to selection based on phenotypic values alone. By integrating pedigree and phenotypic information, BLUPs enable the estimation of breeding values for key commercial traits (TCH, CCS, and fiber), which are then combined into a selection index to rank potential parents for crossing (Wei et al., 2022). While pedigree‐based BLUP selection can more accurately capture the additive genetic merit of related individuals, truncation selection of the highest‐ranking candidates may disproportionately favor closely related parents, thereby increasing coancestry and reducing genetic diversity over successive breeding cycles. Because of these inherent limitations, alternative methods of selection are required to improve genetic gain while preserving the already limited genetic diversity in sugarcane breeding programs. GS IN SUGARCANE: PROMISES AND PITFALLS Principles of GS GS has been a successful strategy in accelerating genetic gain in many plant and animal breeding programs. The approach was first proposed by Meuwissen et al. (2001) and uses dense genome‐wide markers to estimate the genetic value of individuals. GS models the relationship between marker genotypes and trait performance within a training population, which is used to predict genomic estimated breeding values (GEBVs) of untested individuals. Selection candidates are ranked, and high‐value individuals are used as future parents and/or for release as commercial varieties. Thus, GS accelerates gain by reducing the breeding cycle through early selection and crossing of parents and by improving the accuracy of breeding values estimated from early‐stage trials with limited replication. In sugarcane, the use of high‐throughput genomic tools and genetic mapping has begun to unravel the complexities of its polyploid genome. Globally, this transition was accelerated by the development of high‐density SNP arrays, such as the Affymetrix Axiom 345K SNP chip, utilized within Australian and Brazilian breeding programs (Aitken et al., 2016). These platforms provided the first robust mechanism for capturing genome‐wide variation in a high‐throughput manner. A custom 58K SNP array derived from this larger platform has been pivotal in enabling the deployment of technologies such as GS, supporting the large‐scale, high‐throughput genotyping of thousands of clones annually. Subsequently, integration of next‐generation sequencing technologies has broadened the scope of sugarcane research by enabling de novo variant discovery across the entire genome. Genotyping‐by‐sequencing approaches are highly efficient for discovering large numbers of SNPs without the need for a pre‐existing reference genome, thereby capturing more genetic variation underlying important traits (Manimekalai et al., 2020). Targeted sequencing platforms like Flex‐Seq have emerged as the preferred method for validating markers associated with sucrose accumulation and disease resistance (Jaimes et al., 2024). Furthermore, continued development and mapping of the high‐quality sugarcane reference genome (R570) have identified loci and potential molecular targets associated with key traits such as brown rust resistance (Healey et al., 2024). Collectively, these advances enhance the resolution and accuracy of genomic analyses, thereby strengthening the implementation of GS and accelerating genetic gain. Genomic prediction in sugarcane In sugarcane, GS has shown moderate prediction accuracies for key commercial traits such as cane yield and sugar content (Deomano et al., 2020; Gouy et al., 2013; Hayes et al., 2021; Voss‐Fels et al., 2021; Yadav et al., 2021). Deomano et al. (2020) used GBLUP to predict CCS and TCH in a commercial sugarcane breeding population, reporting moderate prediction accuracies of between 0.25 and 0.45. Similarly, Hayes et al. (2021) combined pedigree and genomic information and achieved moderate to high prediction accuracies for CCS, fiber content, and flowering traits in their best model (0.30–0.44). However, they reported low prediction accuracies for TCH, attributing this to its poor narrow sense heritability and strong influence of non‐additive effects. Recently, a study comparing 22 statistical models for genomic prediction, including parametric, nonparametric, and machine‐learning methods, yielded prediction accuracies of 0.36–0.45 for sugar yield and 0.34–0.38 for cane yield (Shahi et al., 2025). Authors attribute this moderate predictive ability on low trait heritabilities and complex genetic architecture. Combined, these results illustrate that GS only captures a portion of genetic variation in sugarcane. Integrating high‐throughput phenotyping to boost genetic gains Advances in high‐throughput phenotyping in sugarcane, particularly the use of unmanned aerial vehicles, present a powerful opportunity to exploit indirect traits with strong genetic correlations to poorly heritable and difficult to measure traits like cane yield (Luo et al., 2023). For example, an indirect selection index incorporating several traits, including canopy temperature, canopy height, and canopy cover, showed high heritability and strong genetic correlations with cane yield, with authors highlighting its superiority over conventional phenotypic selection (Natarajan et al., 2019). Such approaches can improve phenotypic accuracy, which is especially beneficial in early selection stages where competition and environmental effects are most profound, whilst also reducing labor demands associated with measuring high‐biomass crops like sugarcane. Given that GS accuracy depends heavily on phenotyping quality, integrating high‐throughput phenotyping enables the generation of larger, high‐quality training populations, thereby improving genomic analyses and prediction outcomes. Although widespread adoption of such technologies has not yet been realized, the field is rapidly advancing and offers strong potential to enhance breeding outcomes when integrated alongside high‐throughput genotyping methods. Modeling non‐additive effects Given sugarcane's polyploidy and strong non‐additive effects, studies have explored models that include dominance and epistasis to predict clonal performance (i.e., the total genetic value). In clonally propagated crops like sugarcane, this is particularly beneficial, as it can facilitate the rapid identification of superior clones for potential commercial release, as non‐additive effects through advantageous allelic combinations can be preserved. Yadav et al. (2021) modeled dominance and epistatic interactions in an extended GBLUP model and reported low to moderate prediction accuracies for CCS, fiber content, and TCH. They reported a 31% increase in prediction accuracies compared to an additive‐only model, highlighting a substantial improvement for traits heavily influenced by non‐additive variance. However, the authors cautioned that dominance models may overestimate dominance variance and confound additive effects if not properly parameterized. Nevertheless, while useful at predicting clonal performance, non‐additive effects do not contribute to additive genetic gain across generations and cannot be used to select candidates to parent the next generation. Consequently, the focus must remain on the additive component to achieve long‐term gains. Limitations of GS: Risk of diversity loss Despite its advantages, GS risks the rapid depletion of genetic diversity over time. Although beneficial in the short term, selecting high GEBV individuals can rapidly accumulate inbreeding in a population, as elite individuals tend to share the same alleles and are likely related to each other (Sonesson et al., 2012). This loss is further aggravated when coupled with high selection intensities like those used in sugarcane and other plant breeding programs (e.g., Lin et al., 2017). In dairy cattle, where GS was first introduced, the approach initially revolutionized milk yield improvement. Yet today, several populations show alarming signs of genetic bottlenecks, with inbreeding rates climbing faster than anticipated (Makanjuola et al., 2020; Scott et al., 2021; Tenhunen et al., 2024). For crops with a narrow genetic base and high levels of relatedness between parents, like sugarcane, alternative selection strategies are critical to prevent the loss of genetic diversity that will be vital to respond to future challenges (Lin et al., 2017). OCS offers a strategy that maximizes genetic gain under an inbreeding constraint, creating an optimized trade‐off between GEBVs and relationships of selection candidates. OCS: CONCEPT AND RATIONALE Origins and theory OCS is a quantitative breeding approach, which determines the optimal genetic contributions of each candidate to the next generation by weighting their selection not only on genetic merit but also on their genetic relationships to others (Meuwissen, 1997; Sonesson & Meuwissen, 2000; Woolliams et al., 2015). It aims to maximize expected mean breeding value of progeny while constraining average coancestry (and thus inbreeding) of the selected set. This allows breeders to achieve more gains without exceeding acceptable levels of relatedness. The concept dates back to work by Wray and Goddard (1994) and Meuwissen (1997) in animal breeding. Rather than simply choosing a fixed number of “best” parents and mating them arbitrarily, OCS determines how much each candidate should contribute to the next generation, represented as the proportion of offspring it should produce. These contributions are optimized so individuals with high breeding values are favored, but closely related individuals are penalized, ensuring the overall relatedness stays within the target limit. By explicitly managing the trade‐off between gain and diversity, OCS supports sustainable genetic progress across breeding cycles. Mathematically, OCS solves an optimization problem in the form: where ci is the vector of parental contributions, g is the vector of estimated breeding values (EBVs or GEBVs), A is the relationship matrix, and Γ is the desired maximum average coancestry (relatedness). The upper bound cmax can be used to prevent any one individual from contributing excessively. By adjusting Γ and cmax, the breeder controls the trade‐off between short‐term gain and long‐term diversity (Meuwissen, 1997; Woolliams et al., 2015). The OCS framework provides continuous re‐optimization at each generation as new information is acquired. This enables OCS to adapt to changing allele frequencies and inbreeding levels over time. Additional constraints can be included, such as minimum contributions from certain subpopulations or sexes, specific/group matings, selfing, and management considerations (Gorjanc & Hickey, 2018; Kinghorn, 2011; Wellmann, 2019). Moreover, OCS can be extended to accommodate higher ploidy levels, thereby enhancing its applicability to polyploid crops like sugarcane (Peixoto et al., 2025). Success stories of OCS in other crops and species OCS has its strongest theoretical and practical foundation in animal breeding, where it provides an established framework for allocating parental contributions while constraining the expected increase in inbreeding. This is especially important in cattle breeding because intensive use of a small number of high‐merit sires can rapidly increase population coancestry and reduce the variation available for future selection. Extensive animal‐breeding research has shown that OCS can balance improvement in breeding‐goal traits with the maintenance of genetic diversity, while practical tools such as MateSel and optiSel support its application in selection and mating decisions (Kinghorn, 2011; Wellmann, 2019; Woolliams et al., 2015). This history means that OCS is an established breeding framework that now requires adaptation to the biological and operational features of a perennial, clonally propagated, and highly polyploid crop. Tools such as MateR are now beginning to extend this established OCS framework to polyploid crop systems by explicitly accounting for ploidy level and non‐additive (dominance) effects (Fernández‐González et al., 2026). Evidence from crop breeding varies in the degree to which OCS has been implemented and validated. The clearest example of realized application is in spring canola, where Cowling et al. (2023) applied OCS during three cycles of family selection conducted from 2016 to 2020, with multi‐environment field testing in Australia and Canada. Grain yield increased by 2.9 to 4.3% per year, depending on how genetic gain was estimated, while the mating design maintained low parental coancestry and incorporated yield stability and other economically important traits. In common bean, Saradadevi et al. (2021) applied OCS to genomic predictions derived from 33 field trials and identified crosses predicted to increase grain yield by 12.4%, reduce cooking time, and improve iron and zinc concentrations at an achieved coancestry of 0.032. Building on this work, their recent study assessing the realized gain from these crosses confirmed a reduction in cooking time of 8 min on average per year (or 11% per year reduction) (Saradadevi et al., 2026). Together, these studies represent an empirical, data‐informed crossing design with reported realized gains across completed selection cycles, providing empirical evidence for the effectiveness of OCS in crop systems. Research in perennial and clonally propagated species provides a closer biological comparison with sugarcane, although much of this evidence remains based on modeling or retrospective breeding data. In black poplar, Tiret et al. (2021) applied OCS to phenotypic and genomic data from approximately 1000 individuals in the French Populus nigra breeding population. Their modeling showed that both parental contribution and mating structure affected the selection plateau and that compensatory mating could accelerate gain without unnecessarily reducing future diversity. In perennial oil palm, simulations calibrated using genetic parameters from the Deli and La Mé breeding populations showed that shortened genomic‐selection cycles could increase annual genetic gain but also accelerate inbreeding (Seyum et al., 2022). Building on this, Tchounke et al. (2023) identified OCS and mate selection as appropriate methods for explicitly managing long‐term diversity. These studies are relevant to sugarcane because they address long generation intervals, highly heterozygous parents, restricted founder populations, and the need to retain diversity over repeated cycles of selection. However, they should be described as case‐study modeling rather than evidence of multi‐cycle OCS implementation in operational breeding programs. Related simulations in clonally propagated crops have also examined how parent and cross selection should account for heterozygosity, dominance, and inbreeding depression. Using strawberry as an example, Werner et al. (2023) showed that genomic prediction of cross performance could improve long‐term response relative to parent selection based only on GEBVs. Their results also indicated that OCS would be needed when several rapid crossing cycles are conducted without frequent updating of the prediction model. Comparable concerns occur in cassava, where inbreeding depression and the maintenance of heterozygosity affect both parental improvement and clonal performance (Wolfe et al., 2021). Although these studies did not report operational implementation of OCS, they demonstrate that gain‐diversity optimization is especially important in outbred and clonally propagated crops, where additive gain must be accumulated without eroding the heterozygosity supporting commercial clone performance. OCS also has an important role in pre‐breeding and genetic‐base broadening. Cowling et al. (2017) used stochastic modeling of a field pea gene bank to show that OCS could improve exotic and elite germplasm while retaining more diversity than truncation selection. Allier et al. (2020) extended this principle to maize by modeling the coordinated selection of donor‐to‐elite and elite‐to‐elite crosses. Their results showed that OCS could improve donor populations before introgression and manage the gradual introduction of favorable diversity into an elite breeding pool. Together, these studies indicate that OCS can support both recurrent improvement within the sugarcane breeding population and the controlled introduction of diverse Saccharum and related germplasm. OCS IN SUGARCANE Two‐part breeding program OCS could potentially be integrated within a two‐part breeding program to yield sustainable genetic gains under acceptable inbreeding levels. In two‐part breeding programs, the population is divided into (1) population improvement focusing on long‐term additive gain and diversity management and (2) variety development for commercial release (Gaynor et al., 2017). For example, GS could be used to advance the variety development pipeline through the selection and crossing of high GEBV parents. A minimum number of resulting seedlings could then be selected on their predicted clonal performance (PCP), or total genetic value, into later stages of selection (CAT stage). Simultaneously, high GEBV seedlings from these targeted crosses could be selected and introduced back into the parent population, thus contributing to population improvement (Figure 2). Although this approach could be efficient at identifying genetically superior clones, coancestry is not considered when importing candidates into the parent population, which may risk increasing levels of inbreeding. Proposed two‐part Australian sugarcane breeding program with genomic selection. The program is divided into two parts. The variety development pipeline involves selection of parents with high genomic estimated breeding values (GEBVs) to generate large numbers of seedlings. Predicted clonal performance (PCP) of progeny can be used to select a minimum number of candidates to progress to the clonal assessment trial (CAT) stage. Simultaneously, progeny with high‐value GEBVs are introduced back into the parent population through the population improvement pipeline. FAT, final assessment trial; PCP, predicted clonal performance. To balance inbreeding and genetic gain, OCS can be used to determine the optimal contributions of candidate parents for crossing. In this approach, contributions are allocated across all clones in the parental population to maximize expected genetic gain whilst constraining inbreeding under a breeder‐defined threshold. Selection on PCP of the resulting seedlings can then be used to progress a minimum number of candidates through selection (variety development). Optimal contributions of resulting seedlings are calculated by accounting for coancestry among both siblings and the existing parental population (to avoid excessive relatedness), allowing for the identification of progeny suitable for introduction back into the breeding program (population improvement) (Figure 3). Proposed two‐part Australian sugarcane breeding program with optimal contribution selection (OCS). The program is divided into two parts. The variety development pipeline involves selection and crossing of parents with optimized contributions that maximize genetic gain under appropriate levels of inbreeding. Predicted clonal performance (PCP) of seedlings is used to select a minimum number of candidates to progress to the clonal assessment trial (CAT) stage. Simultaneously, optimal contributions of resulting seedlings are calculated, considering the coancestry between siblings and the pre‐existing parent population. Suitable clones are then introduced back into the parent population through the population improvement pipeline. FAT, final assessment trial. Studies have reported larger long‐term genetic gains on candidates selected using OCS in simulated crop and livestock breeding programs (Gorjanc et al., 2018; Pocrnic et al., 2023; Villiers et al., 2024). A simulated two‐part wheat breeding program using optimal cross selection and recurrent GS achieved 78% more long‐term gain than truncation selection in a small population of parents with four cycles per year (Gorjanc et al., 2018). The authors attribute the observed improvement in gain to a reduced loss of genetic diversity and smaller decline in genomic prediction accuracy across cycles, yielding a two‐ to fourfold increase in efficiency compared with truncation selection. Due to the long vegetative evaluation phase of sugarcane and low recombination opportunities, a possible way to implement OCS would be through overlapping cycles. This enables clones from a given cycle to advance through the variety development pipeline, whilst simultaneously allowing candidates in early evaluation stages to be selected via OCS and reintegrated within the population development pipeline for subsequent OCS‐managed crossing (Figure 3). Not only may overlapping cycles reduce generation intervals, early selection of optimized parents and OCS to mitigate crossing of closely related individuals could offer a strategic pathway for long‐term gains. Ancestral germplasm and introgression Expanding this two‐part approach to include pre‐breeding populations can boost diversity in bottlenecked parent populations. Diverse and ancestral sugarcane hybrids often harbor robust alleles conferring benefits in disease resistance and environmental adaptability. However, introgression of ancestral germplasm is often complicated by the presence of linkage drag, which carries unfavorable alleles, such as low sugar content and poor yield. Using OCS to increase genetic gain whilst conserving valuable allelic diversity could narrow the performance gap among modern elite hybrids. By maintaining broader genetic diversity and enabling continued recombination, OCS may indirectly assist breeding efforts aimed at breaking undesirable genetic linkages and reducing the long‐term impacts of linkage drag. By assigning low value clones a small proportion of contributions, OCS can maintain these alleles within the population until, after many generations, they recombine into high‐performing haplotypes. This approach avoids the exclusion of low‐performing ancestral germplasm whose valuable alleles are in linkage with undesirable alleles. Instead, this strategy preserves their broad genetic base while improving access to favorable alleles for effective introgression. EXTENDING OCS TO ADDRESS CHALLENGES IN SUGARCANE Harnessing non‐additivity through mate allocation OCS operates by optimizing the additive component parents should contribute to the next generation. Fortunately, in clonal crops, non‐additivity can be preserved through vegetative propagation. Thus, using strategic matings, favorable allelic interactions, and SCA can be exploited to maximize total genetic value of progeny. Several studies have used GS‐based strategies based on predicted cross performance (i.e., PCP). A stochastic simulation in strawberry found that selection based on genomic predicted cross‐performance had faster total genetic gain compared to selection on breeding values (Werner et al., 2023). Similar findings have been reported in pigs for various growth traits (González‐Diéguez et al., 2019). In sugarcane, mate allocation strategies can exploit the known non‐additive effects underlying commercial traits. Yadav et al. (2023) demonstrated an increase in PCP for TCH, CCS, and fiber by 57%, 12%, and 16%, respectively, compared to breeding value‐based approaches. Such strategies are particularly valuable in clonal crop development, as they enable breeders to identify promising pairings that preserve the favorable allelic combinations underlying superior performance. OCS followed by mate allocation has been successfully applied to identify beneficial cross combination that maintain acceptable limits of inbreeding. For instance, in a pine and eucalyptus tree breeding program, this approach was used to generate hypothetical progeny as a proxy for assessing how specific matings would influence future coancestry (Kerr et al., 2015). Because breeding cycles in tree species typically span 15–20 years, each selection decision carries substantial long‐term consequences, highlighting the effectiveness of this approach in species with long generation intervals. Using the same approach, Hallander and Waldmann (2009) observed that the ranking of tree contributions differed between models considering only additive effects and those incorporating both additive and dominance effects. More recently, Waldmann (2025) demonstrated a two‐stage approach using OCS and mate allocation, highlighting the benefits of using genomic information to derive more accurate measures of true genetic relatedness. Together, these findings demonstrate how this approach can be used to identify superior genetic combinations, reducing resources spent on poor pairings. In sugarcane, combining OCS with mate allocation offers an approach that exploits both additive and non‐additive objectives. The former maximizes the additive genetic value of parents under an inbreeding constraint, whilst the latter exploits the SCA of favorable alleles during cross planning. Thereby, clones with high additive merit can be recycled back into the population whilst simultaneously harnessing their non‐additive benefits for rapid development of elite varieties, all whilst remaining under acceptable limits of inbreeding (Figures 3 and 4). Extensions on the optimal contribution selection (OCS) framework for sugarcane. Methodology that addresses the challenges in sugarcane breeding, the additional information they incorporate, and potential improved outcomes. Modeling allele dosages for accurate prediction Modeling allele dosage effects offer a more accurate representation of genotype class effects and underlying genetic architecture (Garcia et al., 2013; Osborn et al., 2003). However, allele dosage studies in polyploid species have contrasting findings. In tetraploid potato, authors reported improved prediction accuracies compared to standard diploid models for some traits (Endelman et al., 2018). However, in tetraploid blueberry no significant difference was identified between methods (De Bem Oliveira et al., 2019; Ferrão et al., 2021), suggesting that allele dosages behave differently across polyploid species. In sugarcane, many genomic prediction models apply a pseudodiploid method, as traditional SNP arrays often produce ambiguous allele calls (0/1/2), recording markers as present/absent. Consequently, this simplification ignores dosage variation and potentially limits predictive ability. A recent study modeling ploidy in sugarcane reported improved prediction accuracies in commercial traits compared to discrete parameterization (Yadav et al., 2024). Yadav et al. (2024), using a continuous genotype method, reported a 5%–7% increase in prediction accuracy for CCS and fiber but saw no improvement in TCH. They attribute this limited benefit to the prevalence of low heterozygosity in the population, suggesting many alleles were already fixed by selection. Another study showed similar findings, reporting no significant improvements in prediction accuracy for yield traits (Batista et al., 2022). They too identified low heterozygosity within their small commercial breeding population, reporting over 50% had allele dosages of zero or one. These findings suggest that applying allele dosage modeling may yield larger benefits in pre‐selection populations or those that retain higher heterozygosity (less inbred). Polyploid‐specific genomic relationship matrices can better capture genetic relationships among individuals compared to diploid approximations. They can account for both additive effects and allele dosages to capture the sharing of alleles (and their copies) for better estimates of coancestry and inbreeding. Using these matrices in OCS will enable consideration of allele dosage differences when determining contributions. Moreover, accounting for these can aid in preserving more gene copy diversity, maintaining a broad genetic base, which avoids high levels of homozygosity. For instance, OCS with allele dosages in sugarcane can penalize selection of clones that are homozygous for many loci and prioritize selection on clones with more heterozygosity (within and across loci). Here, additive gain is optimized such that heterozygosity is preserved (Figure 4). However, implementing dosage‐aware OCS in sugarcane remains challenging due to the species' highly polyploid and aneuploid genome, variable ploidy levels among loci, and the difficulty of accurately estimating allele dosage from routine genotyping platforms (as discussed above). Consequently, incorporating allele dosage information into OCS should be regarded as a high‐priority research area rather than a mature breeding technology. Continued improvements in genome assemblies, haplotype resolution, and dosage calling will be essential before the full benefits of dosage‐aware OCS can be realized in sugarcane breeding. Multi‐species germplasms to capture the history of modern sugarcane breeding Research has shown that ancestry‐specific differences can influence multi‐species genomic analyses. This creates a challenge, as subspecies may not share the same genetic patterns due to mutations independently arising within the two (Warburton et al., 2021, 2023). For instance, mutations that arise from a common ancestor may have become fixed in one but not the other subspecies due to drift (Bolormaa et al., 2013; Kemper, Hayes, et al., 2015). If not accounted for, these “ghost‐quantitative trait loci” (where one SNP is present in one population but not the other) can have differing effects on complex traits, impacting the accuracy of genomic analyses (Kemper, Reich, et al., 2015). Several studies in livestock have demonstrated marked differences in the SNP effects between subspecies and report improvements in prediction accuracy when these differences are accounted for (Bolormaa et al., 2013; Duenk et al., 2019; Warburton et al., 2023). For example, using this approach, Warburton et al. (2023) reported less biased and more accurate multi‐species genomic evaluations when breed‐of‐origin was accounted for in composite tropical cattle. To capture these effects in sugarcane, separate marker (or haplotype) effects should be assigned based on its species of origin. OCS can incorporate ancestry considerations by imposing constraints on the proportion of genome from each species. For instance, bounding the contributions of S. spontaneum through minimum and maximum thresholds can help preserve valuable alleles whilst preventing the decline in sugar content associated with excessive S. spontaneum introgression (Figure 4). OUTLOOK AND CONCLUSION Sugarcane provides most of the world's sugar and biofuel consumption; however, the challenges of polyploidy, long‐breeding cycles, and complex genomes have resulted in stagnating gains. Here we propose the conceptual framework of OCS to address these challenges, representing a balancing trade‐off between genetic gain and diversity. Genetic diversity management strategies, like OCS, are widely regarded as the gold standard in livestock breeding programs, enabling sustained genetic improvement while conserving the genetic diversity required to support long‐term and sustainable genetic gain. Whilst this approach remains conceptual in sugarcane specifically, it is grounded in theoretical principles supported by empirical evidence from other crops and livestock systems. When combined with GS and extended modeling of non‐additive effects, OCS is expected to accelerate sugarcane improvement whilst avoiding a decline in diversity that threatens long‐term gains. Integration of OCS with such genomic tools and methods requires careful consideration and investment. For instance, high‐quality phenotypic and genotypic data to provide accurate representation of the genetic merit and relatedness of individuals is required. Furthermore, breeders must choose appropriate inbreeding constraints that are consistently updated over time based on population monitoring. However, several key areas remain where improvements in technology could further strengthen the effectiveness of OCS in sugarcane: Improved genomic technologies and tools: Combining multiple genomes into a pangenome for sugarcane may provide additional information on variation in genome structure, such as allele dosages, which can be used to improve prediction outcomes. A sugarcane pangenome consisting of nine assemblies across four different Saccharum species was recently developed, capturing 82% of sugarcane genomic diversity (compared to ∼34% when using a single reference genome), with authors highlighting the benefits to allele variant discovery (Huang et al., 2026). Thus, integrating such technologies may facilitate the identification of novel allelic variants and allele dosage effects influencing key traits. Improved genomic technologies and tools: Combining multiple genomes into a pangenome for sugarcane may provide additional information on variation in genome structure, such as allele dosages, which can be used to improve prediction outcomes. A sugarcane pangenome consisting of nine assemblies across four different Saccharum species was recently developed, capturing 82% of sugarcane genomic diversity (compared to ∼34% when using a single reference genome), with authors highlighting the benefits to allele variant discovery (Huang et al., 2026). Thus, integrating such technologies may facilitate the identification of novel allelic variants and allele dosage effects influencing key traits. Polyploid‐specific OCS and GS software: Tools that handle the specific intricacies of polyploid genotypes such as allele dosages and multi‐species ancestry are needed. Machine learning and Bayesian modeling approaches may improve cross‐performance prediction in polyploid contexts. Enhanced genotyping platforms could enable more precise interpretation of dosage information. Haplotype‐based breeding could combat linkage drag associated with introgression. Polyploid‐specific OCS and GS software: Tools that handle the specific intricacies of polyploid genotypes such as allele dosages and multi‐species ancestry are needed. Machine learning and Bayesian modeling approaches may improve cross‐performance prediction in polyploid contexts. Enhanced genotyping platforms could enable more precise interpretation of dosage information. Haplotype‐based breeding could combat linkage drag associated with introgression. Simulation studies tailored to sugarcane: By accounting for ploidy, genome size, trait architecture, and breeding cycle length, optimal inbreeding rates and contributions can be determined. Digital twins can model OCS under different breeding scenarios, offering a customized approach. Simulation studies tailored to sugarcane: By accounting for ploidy, genome size, trait architecture, and breeding cycle length, optimal inbreeding rates and contributions can be determined. Digital twins can model OCS under different breeding scenarios, offering a customized approach. Proof‐of‐concept studies: Pilot implementation within breeding programs is required to quantify its impact on genetic gain, diversity, and operational feasibility. Proof‐of‐concept studies: Pilot implementation within breeding programs is required to quantify its impact on genetic gain, diversity, and operational feasibility. Improved phenotyping and prediction accuracy: Advances in high‐throughput phenotyping and genomic prediction could improve breeding value estimation, thereby enhancing OCS outcomes. Improved phenotyping and prediction accuracy: Advances in high‐throughput phenotyping and genomic prediction could improve breeding value estimation, thereby enhancing OCS outcomes. Sugarcane breeding must balance short‐term genetic gain with the maintenance of long‐term genetic diversity. OCS provides a systematic framework to achieve this balance by simultaneously optimizing parental contributions and constraining relatedness. Although further methodological and technological developments are required, OCS represents a promising strategy for sustaining future improvements in yield, quality, and resilience. AUTHOR CONTRIBUTIONS Natalya Vo Van‐Zivkovic: Visualization; writing—original draft; writing—review and editing. Ben J. Hayes: Conceptualization; supervision; writing—review and editing. Karen Aitken: Supervision; writing—review and editing. Eric Dinglasan: Supervision; writing—review and editing. Terry Morgan: Supervision; writing—review and editing. Seema Yadav: Conceptualization; supervision; writing—review and editing. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest. ACKNOWLEDGMENTS This research was supported by the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Wilmar Sugar Australia Limited, and Sugar Research Australia. All figures were created in BioRender by Vo Van‐Zivkovic (Figure 1: https://BioRender.com/50zah6d; Figure 2: https://BioRender.com/zc7tjs4; Figure 3: https://BioRender.com/wy5j1yl; Figure 4: https://BioRender.com/9g5cj6q). Open access publishing facilitated by The University of Queensland, as part of the Wiley ‐ The University of Queensland agreement via the Council of Australasian University Librarians Vo Van‐Zivkovic, N. , Hayes, B. J. , Aitken, K. , Dinglasan, E. , Morgan, T. , & Yadav, S. (2026). Balancing gain and diversity: Optimal contribution selection for sustainable genetic improvement in sugarcane. The Plant Genome, 19, e70314. 10.1002/tpg2.70314 Assigned to Associate Editor Qingyi Yu. REFERENCES