Epistasis

Now for some science.  Epistasis, something my lab group has been interested in for a number of years, is interaction of alleles (the versions of genes that diploid individuals in sexually reproducing species have two of, one from each parent) such that the effects of two (or more) alleles do not equal the sum of the individual effects of each allele.  For example, at two gene loci a and b if there are wild type alleles (a+ and b+) that each contribute 1 unit to a phenotype and two mutant alleles a1 and b1 that each contribute 3 units by themselves (thus the diploid genotypes a+a1 b+b+ and a+a+ b+b1 both have values of 1+3+1+1=6), if there were no epistasis then the genotype a+a1 b+b1 would be expected to have a phenotypic value of 1+3+1+3=8.  However, it is often found in real biology that the effects of two mutant alleles (or two alternative wild type alleles) do not precisely add up to the sum of their individual effects, such that a+a1 b+b1 might have a genotype of 10 (synergistic epistasis) or 6 (masking epistasis; there are other terms for this that are escaping me at the moment).
If this seems complicated now, it gets a lot more complicated, especially when more than two loci are involved. Population genetics theory has been done on positive and negative epistasis for deleterious alleles.  Positive epistasis occurs when as deleterious mutations accumulate, they have their worst effects when there are few of them in a genotype, with decreasing magnitudes of effect as more are added (this resembles the masking case above).  Negative epistasis occurs when the effect of adding another deleterious mutation gets worse and worse as more are added (this resembles the synergistic case above).  There turns out to be little empirical evidence of the latter in nature (or at least when genetic experiments have looked for it).  However, in theory negative epistasis is required for sexual reproduction to evolve in an ancestrally asexual species (i.e. for sexual reproduction to confer an evolutionary advantage greater than its costs).  This latter point is something I read about today (Peck JR, Waxman D, Welch JJ. 2012 Hidden epistastic interactions can favour the evolution of sex and recombination. PLoS One 7:e48382) and am still trying to wrap my mind around.
In general, epistasis determines the effect of a mutation. This depends on the "genetic background" (the genotypes at all of the other loci in the genome). In a population, there is typically a near- infinite number of potential genetic backgrounds, and some very much smaller number of predominant types that are largely determined by the major genes that interact at the molecular level with the gene of interest (either at the DNA, RNA, or protein level).  This becomes very important when trying to find and decipher the effects of disease genes in humans.