The existence of LD means that we do not actually have to genotype the causal locus to detect association; we simply have to genotype variation very near the causal locus. In practice, it is exceedingly difficult to determine whether an association signal results due to direct or indirect association, though the default assumption must be that it is indirect. First, most genetic variants (like the alleles of most markers we study) have no obvious functional significance. Only ∼2% of the genome represents genes, and the exons (segments of coding information) are interspersed with often much larger introns (segments within the gene that do not carry coding information) so only ∼1% of the DNA in the genome actually codes for functional molecules. Thus, the chance of a variant impacting the function of a key molecule is small. Second, this difficulty is compounded by the expectation that most of the alleles influencing traits common in the population are common and of relatively small effect (i.e., account for only small quantities of the total trait variance). Such alleles are unlikely to have major
