A couple new papers review the factors that play a role in determining an individual's epigenotype and the role of said epigenotype in the aetiology of autism spectrum disorders.

We still do not understand the rules governing the establishment and maintenance of the epigenotype at any particular locus. The underlying DNA sequence itself and the sequence at unlinked loci (modifier loci) are certainly involved. Recent support for the existence of transgenerational epigenetic inheritance in mammals suggests that the epigenetic state of the locus in the previous generation may also play a role. Over the next decade, many of these processes will be better understood, heralding a greater capacity for us to correlate measurable molecular marks with phenotype and providing the opportunity for improved diagnosis and presymptomatic healthcare.

Here's a question that's been bothering me: how could one demonstrate the extent of epigenetic inheritance in humans? Any "easy" look at heritability is confounded by genetic effects. Here's my experiment: I'd need a number of genetically identical sperm with different epigenetic profiles and a number of genetically identical eggs with different epigenetic profiles (and assume I know these genome-wide profiles). I make me a bunch of twins, and determine their genome-wide epigenetic profile at some stage of development. Any correlation between the epigenetics of the children and the epigenetics of the parents would be most parsimoniously explained by epigenetic inheritance. This is probably both technically and ethically impossible, so is there any other way?