Associating of causal polymorphism with complex phenotypes can provide a better understanding of the mechanisms underlying developmental and biochemical constraints, thus enabling accelerated crop improvement. Advances in genotyping technologies and computational approaches have reshaped the way genetic analysis is conducted and have allowed capturing a substantial proportion of the global genetic diversity for genome-wide association studies 
. While different “association mapping” approaches in plants are focused mainly on additive variation between inbred lines, much less attention has been paid to developing experimental and computational approaches for identifying and deciphering the possible role of intra-locus interactions that contribute to the phenotypic landscape of the hybrids. The underlying genetics of heterosis has long been debated–ever since it was first observed and documented by Charles Darwin 
and later studied experimentally with natural and artificial populations of various organisms. Although it is agreed that increased homozygosity often lowers fitness-related characters (survival, growth rate and fertility), at the heart of the debate is the extent to which this can be attributed to increased homozygosity for partially recessive detrimental mutations (dominance), rather than changes in homozygosity for alleles at loci with heterozygote advantage (overdominance model 
). This debate also holds for agricultural yield: is the vigor of the hybrids the outcome of many dominant loci with intermediate effects working in a multiplicative manner on different yield-associated traits, or are those overdominant loci key regulators in several pathways throughout plant development?
Two recent reports illuminate this debate, although the final conclusion of both may still be linked to the type of plant system and phenotype analyzed, as well as to the genetic background (whole genome hybrids compared to isogenic homozygous background). Riedelsheimer et al
crossed 285 diverse Dent inbred lines from worldwide sources with two testers and predicted their combining abilities for seven biomass- and bioenergy-related traits using 56,110 sngle-nucleotide polymorphisms and 130 metabolites. Then they performed genome-wide association scans using a Q+K model 
and found no strong association signals, even though population size, heritabilities of the traits, extent of linkage disequilibrium and marker density were sufficiently high to detect large quantitative trait loci (QTLs). Based on these results, it was concluded that the genetic architecture that underlies heterosis is close to an infinitesimal model. On the other hand, Krieger et al
searched for genes that cause heterosis in tomato (Solanum lycopersicum)
by crossing 33 diverse fertile mutants with the matching non-mutagenized parent known as ‘M82’ to create isogenic mutant heterozygotes. Comparison of their yield traits identified six mutant heterozygotes that showed yield heterosis, with the individual effects ranging from 36% to 88%.
To date, no individual overdominant locus has been isolated by unbiased QTL mapping and recent attempts to zoom in on genomic loci associated with heterosis have mostly made use of biparental populations. This is performed by estimating the heterosis phenotype through test-crossing introgression lines or recombinant inbred lines to the recurrent parent 
and linking the heterotic mode of inheritance with marker-defined genetic intervals. Additional approaches have included the use of immortalized F2 populations 
and the fine-mapping of heterotic QTLs in nearly isogenic lines 
. Lariepe et al
recently reviewed the use of biparental populations for dissecting heterosis and extending the use of the North Carolina III (NCIII) design with markers 
to develop and implement a multiparental-connected model. This allowed studying heterosis in families derived from both related and unrelated parents, and comparing not only contrasts between homozygous and heterozygous genotypes, but also contrasts between heterozygous genotypes at each locus 
. One approach for studying heterosis, devised very early in plant and animal breeding, was generation of a cross matrix (diallel) between founder lines (FLs) followed by phenotypic analysis of these inbred lines and their hybrids. Kearsey 
was the first to compare the merits of five designs–full diallel, half diallel, partial diallel, NCI and NCII. These studies, which were conducted before the molecular concepts of genetics had been formulated, were concerned with dissecting the heterotic variation into its components (additive, dominant and epistatic), as well as with determining general and specific combining abilities between the different founder inbreds. With the advent of genetic marker technologies, it became feasible to estimate the relationship between overall genetic distance and the magnitude of heterosis calculated by the general combining ability of the parents 
. Cho et al
connected genetic heterogeneity and heterosis at each locus by calculating the differences in the levels of best parent heterosis (BPH) between different genotypes. The ANOVA comparison between BPH values of homozygotes and heterozygotes at each locus concluded that more than 30% of the markers were associated with a heterotic mode of inheritance. Note that this analysis did not distinguish between different allelic combinations among the hybrids and treated the heterozygotes in each locus as a single group.
Previous studies have reported a high level of heterosis in Sorghum bicolor
, the fifth most important crop in the world (www.FAO.org
), which exhibits a consistent yield increase in hybrids vs. varieties (up to more than double) under a wide range of growing and management conditions 
. Interestingly, these early studies led to the identification of one of the few cases of whole-plant heterosis resulting from the heterozygous condition of a single mutation that affects duration of growth 
. However, since the underlying gene was not identified, the possibility that this non-additive mode of inheritance is due to pseudo-overdominance 
of more than one gene still remains open. Heterotic groups are not as clearly defined in sorghum as in maize, and studies using molecular markers have shown that prediction of heterosis is enhanced by using particular linkage groups in models attempting to associate genetic distance and hybrid-group performance 
. Although these observations lend some support to the attractive possibility of identifying single genes underlying heterosis in this major crop plant, the heterosis map of sorghum, which would allow alignment with existing maps in maize 
, rice 
and other plants, is still lacking.
Here we present the first use of heterotic trait locus (HTL) mapping to identify intra-locus interactions mediating an over- or underdominant mode of inheritance. The wide diallel population of S. bicolor is derived from 19 founder lines (FLs) selected from a wide gene pool. This population was phenotyped for 2 consecutive years and statistical and computational tools were integrated to test for association between specific intra-locus interactions and overdominant mode of inheritance. We discuss possible mechanisms underlying heterosis for grain yield in hybrids based on our experimental data and demonstrate that the results of this fast-track mapping are validated by the overdominant mode of inheritance of one HTL in the consecutive F2 population. Finally, we propose the incorporation of individual resequenced genomes of only diallel FLs to zoom in on hitherto unknown loci underlying heterosis, thereby paving the way for an improved understanding of molecular mechanisms underlying this elusive and important phenomenon.