On The Use Of Mathematically-Derived Composite Traits And Their Efficiency In Quantitative Trait Locus Mapping

Composite traits, such as addition, subtraction, multiplication, and division, are indices mathematically derived from two or more individual traits, which have been frequently used by breeders in selection for favorable gene or gene combinations for two or more traits simultaneously. When used in QTL mapping, composite traits often show discrepancy with QTL identified by individual traits. Up to now, there have been few researches concerning the suitability for using composite traits in QTL mapping, reasons for the difference of QTL mapping results from individual and composite traits and the nature of composite-only QTL.A total of 29 putative genetic models classified into 5 groups (genetic model A-independent genetic model, genetic model B-QTL for 2 component traits are independent, genetic model C-QTL for 2 component traits are linked on two different chromosomes, genetic model D-2 QTL for component traitâ… and 1 QTL for component traitâ…¡are linked and genetic model E-4 QTL for both component traits are linked on the same chromosome) and 4 genetic models based on the mapping results in 4 actual maize and rice mapping populations were used to investigate the efficiency of using composite traits in QTL mapping, and to explore the reality of composite-only QTL, i.e., QTL identified by a composite trait but not by any individual trait. The 29 putative genetic models contained independent QTL and QTL linked in various genetic distances. The main results from this study are described as follows.1. When there are two QTL affecting either of the two individual traits, the number of QTL affecting a composite trait could be four. Simulation results of independent genetic model indicated that the detection power of QTL affected component traits was consistently greater than that in other linkage genetic models. In addition, the detection power of QTL affected component traits was greater than that of composite traits. Simulation results of linkage genetic model indicated that the detection power of QTL affected component traits was still greater than that of composite traits. What's more, there was an increasing trend of the detection power of QTL for both component and composite traits with the increasing of the genetic distance between QTL. In both independent and linkage genetic models, much lower detection power and a higher false discovery rate (FDR) were observed when composite traits were used.2. In the actual populations, simulations were designed based on the estimated QTL positions and effects. When composite traits were used, QTL detected by both individual and composite traits had comparable power, but those detected by component traits but not by composite traits had low detection power. In the maize RIL populations, FDR from subtraction and division were much higher than that from addition and multiplication. While in the rice F2 populations, FDR from addition and multiplication were much higher than that from subtraction and division.3. Composite traits had more complicated architecture, as expected. The use of composite traits in QTL mapping increased the gene number, which made it more difficult to control genetic background variation, therefore reduced the efficiency of QTL mapping and phenotypic variation explained by individual QTL. In addition, the use of composite traits in QTL mapping caused higher-order gene interactions than observed in individual traits and complicated the linkage relationship between QTL.The increased complexity of the genetic architecture of composite traits reduces QTL detection power and increases FDR.4. For genetic model A and B, heritabilities for the four composite traits were equal to or lower than those for the two component traits. In genetic model C, D and E, heritabilities of addition and multiplication were higher than those of two component traits; heritabilities of subtraction and division was lower than those of two component traits. What's more, heritabilities of addition and multiplication were higher than those of subtraction and division. Heritability in other four QTL distribution and effect models in the maize and rice population follows a similar trend, except that the change in genetic variance is also caused by pleiotropic QTL in the maize population.5. Composite-only QTL identified in actual mapping populations can be explained either as minor QTL not identified by individual traits or as false positives. Composite traits should be used with caution in QTL mapping. However, the less efficient of using composite traits in QTL mapping should not rule out their use in breeding since they have a great advantage when selecting all favorable genes simultaneously.