Aspects of genetics and statistics in designing marker assisted selection programs for sugarcane.

Population genetic theory provides tools that can be used to examine the efficiency of different marker discovery and utilisation schemes. Projects suited to marker discovery for genomics research will not necessarily be optimal for marker assisted selection (MAS), depending on the frequency of the markers and their linked genes. The efficiency of conventional selection versus MAS must be compared at the correct level of replication, and expressions relating efficiency across different numbers of replication have been derived. The use of selfed populations for marker identification will be more effective than outcrossed populations for some applications. In addition, markers for undesirable traits (e.g. disease susceptibility) may be more useful than markers for positive traits in some circumstances. In sugarcane, MAS is likely to be most useful for identifying rare recombinants for multiple traits. Introduction in heterozygosity over the S. officinarum Designing and implementing effective marker component. assisted selection programs involves issues regarding Gene frequency and segregation of S. spontaidentifying markers and linked QTLs or genes neum markers is likely to change with increasing affecting the phenotype of interest, verifying the generations from the original hybridisation. utility of the identified markers as a selection tool, In sugarcane, inbreeding will also result in an and integrating MAS into breeding programs. Each apparent increase in heterozygosity if this is of these is affected by factors such as the gene fremeasured by molecular marker polymorphism. quency of the markers and QTLs, their linkage relationships, the heritability of the traits concerned, and Trait heritability the biology of the crop. As inbred lines cannot be For MAS to be useful, it must be more effective used in sugarcane breeding, Q T L ~ or genes linked to than ~0n~ent i0nal selection. The efficiency of selecmarkers cannot be incorporated into new varieties tion can be determined from the clonal repeatability without avoiding the effects of linkage drag. These (CR), and will depend on trial design, number of repfactors need to be integrated into a strategy that offers lications etc. The efficiency of MAS can be estimated the best chance of successfully exploiting MAS, to as the proportion of variance accounted for by the ensure that the hours of laboratory work and the costs n~arkers (R2), and must be at least equal to CR for involved have some ultimate benefit. MAS to be beneficial. This will depend on the stage of the selection program at which MAS is to be done. Gene frequency Table 1 shows values of R2 equal to CR at one to four It is well understood that only alleles with a low replications. For a trait with a clonal repeatability of frequency in the population will show segregation in 0.6, markers would need to ascribe 7% of the phenosugarcane, and be useful for mapping or MAS (Wu typic variance to be more effective than conventional et al., 1992). Some of the implications for marker selection, assuming that the selection was done in identification and MAS resulting from this are unreplicated trials. For the same trait measured in summarised below. three replications, markers would need to ascribe 28% of the variance to be effective. MAS is most * As selection results in an increase in gene freeffective for traits with low heritability, or that are quency, many genes linked lraits difficult to measure (van Berloo and Stam, 1998). may not segregate in a useful manner in the These traits, however, will also be difficult to find breeding population. markers for, as the power of QTL detection is As cO-ancestry is quite high in many breeding directly related to trait heritability (Kaeppler, 1997). programs, the chance of finding alleles specific to a few individuals may be low. Table 1-Values of R2 equivalent to values of CR for trials with Although markers will be identified in breeding one to four replications. populations, their usefulness for MAS will depend on their frequency in the breeding population. 0.64 0.56 0.44 0.25 As the S. spontaneum portion of the genome is 0.6 0.36 0.28 0.18 0.07 comparatively low in modern hybrids (D'Hont 0.4 0.16 0.11 0.06 0.02 et al., 1996), it will show an apparent increase 0.2 0.04 0.03 0.01 0.003