Results Obtained from Single Factor ANOVA

The following information is obtained from the ANOVA method of QTL detection:

1. A rough estimate of the QTL position. A QTL is inferred to be located close to the most significant marker within a given chromosome region (i.e., the marker with the lowest P-value or highest R2 value). An example is shown in Table 2. Several markers in the same region were significantly associated with the phenotypic trait, but the authors concluded that umc105a was located closest to the QTL because it had the highest R2 value (see point 4 below).

Marker Location on chromosome 9 P-value R2(%) Parent contributing higher value allele
c1 18 cM <0.0001 6.3 GT119
umc113a 19 cM <0.0001 6.3 GT119
sh1 23 cM <0.0001 6.7 GT119
bz1 24 cM 0.0018 4.8 GT119
umc105a 38 cM <0.0001 10.8 GT119
Wx1 47 cM 0.0005 5.3 GT119

Table 2. Partial ANOVA results from QTL study by Byrne et al. (1996).

2. Measure of statistical significance: P-value. This value indicates the probability of obtaining results this extreme if the marker was not associated with variation for the trait. For example, a P-value of 0.01 indicates a 1% probability that these results would have been obtained in the absence of a marker-trait association. The lower the P-value, the higher the probability that a QTL truly exists in the region of the marker. Many researchers do not have confidence in a QTL unless the P-value of a linked marker is less than 0.01.

3. %R2. This value indicates the relative importance of a QTL in influencing a trait. It is the percent of the total phenotypic variance for the trait that is accounted for by a marker. %R2 is obtained by multiplying the R2 value provided in the ANOVA results by 100.

4. Source of the favorable allele (Parent A or Parent B). Mean values for the marker classes are compared, and the most favorable mean is considered the source of the desired QTL allele. For example, if the mean grain yield of all lines with the ‘A’ marker pattern is 6 tons/ha, and the mean for all lines with the ‘B’ pattern is 3 tons/ha, then Parent A is identified as the source of the favorable allele. Bear in mind that for some traits, such as disease severity, a lower mean value will be preferred.

5. Estimates of additive and dominance effects. The average additive effect of an allele is estimated as

(Mean of A marker class – Mean of B marker class) / 2.

If the A mean = 6 tons/ha and the B mean = 3 tons/ha, then the average additive effect of substituting an A allele for a B allele at that marker is (6 – 3)/2 = 1.5 ton/ha. The difference between means is divided by 2 because the A class (AA genotype) differs by two allele substitutions from the B class (BB genotype). Note that if the A class is bigger than the B class, the additive effect will be positive; if the reverse is true, the effect will be negative.

Dominance effects can be estimated in populations in which heterozygotes are represented, e.g., an F2 population, which has an expected 50% rate of heterozygosity at each marker locus. The dominance effect at a locus is estimated as

Mean of heterozygous (H) class – [(Mean of A class + Mean of B class) / 2].

In other words, the dominance effect is the deviation of the heterozygous condition from the midparent mean. If the H, A, and B classes = 5, 6, and 3 tons/ha, respectively, then the dominance effect = 5 – (6 + 3)/2 = 5 – 4.5 = 0.5 tons/ha.

Limitations of the Single-Factor ANOVA Method

  • It is difficult to know what proportion of the organism’s genome is covered by a set of markers because chromosome maps are usually not constructed.
  • QTL locations are detected only in terms of the nearest marker and, therefore, are imprecisely estimated.
  • The size of the QTL effect is confounded with distance of the QTL from the nearest marker.