Marker-assisted selection for multiple genes

Using markers to select for multiple genes is more complex, and less proven, than selection for a single gene. Population sizes required to recover individuals with all the desired marker patterns increase exponentially with the number of genes involved. In a backcrossing scheme, there may be little opportunity to select for the recurrent parent genome, because so few individuals will have the desired marker pattern at all the target loci. If some of the genes are QTLs, whose locations and effects are often imprecisely estimated, then there is uncertainty that the results of MAS will meet expectations. Finally, the more genes undergoing selection, the greater the chance of incorporating unfavorable alleles through linkage drag.

Some suggestions for using markers to select for multiple genes are as follows:

Figure 9. LOD curve from a QTL analysis, indicating that the most likely QTL position (peak of the curve) is in the middle of a 24 cM marker interval. To select for the favorable alleleat the QTL, selection on the basis of both flanking markers (asg20 and whp1) is advisable.

  • Limit the number of genes undergoing selection to three or four if they are QTLs selected on the basis of linked markers, and to five or six if they are known loci selected directly (Hospital, 2003).
  • Target only verified QTLs that have medium to large effects and that are consistently detected in a range of environments.
  • Examine the QTL analysis results carefully to decide which markers to select, as illustrated in Fig. 9.
  • If desired, an index can be constructed that weights some markers differently than others, depending on their relative importance to the breeder. Flint-Garcia et al. (2003) provide an example of an index they employed to select for QTLs with different effect sizes.
  • Consider a stepwise backcrossing procedure (Hospital, 2003). For example, if four target genes are to be introgressed into the same genetic background, one could first conduct two parallel backcross schemes, each incorporating two target genes. Then the selected individuals from each scheme are crossed and plants with all four targets identified. This procedure gives greater opportunity to conduct background selection for the recurrent parent genome than selecting for all four targets simultaneously.
  • Consider strategies described by Bonnett et al. (2005) for increasing the frequency of target alleles in a population, thereby allowing reduction of the population size needed to attain selection goals. Their strategies, which can reduce minimum population size by up to 90%, include F2 enrichment, backcrossing, and inbreeding.


  • MAS for multiple QTL markers was compared to phenotypic selection in maize by Flint-Garcia et al. (2003). QTLs had previously been identified for 2nd generation European corn borer (ECB) resistance in one population and for rind penetrometer resistance (RPR), an indicator of stalk strength, in three populations. For each trait and population, selection was carried out as indicated in Fig. 10, with the 10 highest or 10 lowest families selected in each fraction. Each of the five selected sub-populations was recombined by random mating the selected families, followed by evaluation in field trials.

Figure 10. Selection scheme for comparing MAS with phenotypic selection for rind penetrometer resistance (RPR) in maize (Flint-Garcia et al., 2003).

  • Results showed that MAS was effective for both traits, but not always as effective as phenotypic selection. In some cases, MAS was effective in moving the population in one direction (e.g., ECB susceptibility), but not in the other. Logistically, MAS was considered more advantageous for ECB resistance than for RPR, because of the greater time and expense required for ECB resistance evaluation.
  • The ability to pyramid multiple resistance genes in the same variety is often mentioned as an advantage of MAS. Castro et al. (2003) provided an example of that strategy in barley by combining a qualitative gene with QTL alleles for resistance to barley stripe rust. Preliminary results indicated combining qualitative and quantitative resistance genes improved resistance levels in the presence of a virulent race of the pathogen.
  • The value of alleles from wild relatives was demonstrated in a MAS study for blackmold resistance in tomato (Robert et al., 2001). Five QTL alleles for resistance, previously detected in wild Lycopersicon cheesmanii, were backcrossed into a cultivated tomato background and the backcross progenies were evaluated. Three of the five alleles were effective in reducing disease severity; however, only one of the effective alleles was not associated with negative horticultural traits. The authors proposed fine mapping studies to determine if markers could be used to separate resistance from the undesirable traits.