I'm skeptical of ScienceDaily over-hype. But the quotes from the researchers seemed reasonable.

It's possibly that methylation adds an extra layer of a 'code' that evolves faster than the nucleotide sequences. I need to read up on methylation to come to a better understanding of what it entails.

One thing that added strength to their argument was the proportion of change in methylation per generation, compared to the total overall possible methylation sites:

This is a key factor in evolution--essentially, the mutation rate of epigenetic variation. Too low, and there's no change, no evolution. Too high, and the changes all become random noise, and any adaptations in one generation are destroyed in the next.

From studying evolutionary algorithms, it is often found, in a wide variety of different models of evolution, that a mutation rate will be quite low, but never zero. Often, models will use mutation rates of 3%, 1%, or lower, per generation. Anything more than 5% is usually too noisy to work. Extremely low mutation rates tend not to go anywhere, they take too much time. But nature has all the time in the world, so low mutation rates are not uncommon in biological life. Zero, however, is out of the question. Zero mutation eventually leads to stagnation and extinction.

Now, they said the rate in this experiment was 'a few thousand' out of 'six million'. That looks very promising for a natural mutation rate:

3,000 / 6,000,000 = 1 / 2,000

= 0.0005 = 0.05%

If they are right, then it looks to me like that mutation rate would support adaptation by natural selection. It's not too high, like 5%, but it's significantly higher than 0% (consider that you can have many thousands or millions or billions of organisms per generation, so such a 'low' rate will actually yield several mutations per generation. That's plenty).

However, again, with the over-hype. Although this would genuinely be a new 'layer' of evolution. It would not overturn our current theory in any way; it would just add more depth to it. You would still have the same five basic concepts working: Generation, Variation, Communication, Selection, Iteration. It's just that the Generation, Variation, and Communication would be working via methylation sites, rather than via nucleotide bases. It would still be evolution by Darwinian Natural Selection.

You still need the DNA code  - all epigenetics can do is switch genes off.

It cannot create new genes, so we still definitely need DNA to evolve in the first place, even if just to give epigenetics material to work with.

Well, Bob, think about it this way. Epigenetics requires DNA, yes, because DNA was first, and the methylation happens on the DNA itself.

However, epigenetics isn't just switching genes on and off, because those switchboard settings also get replicated from one cell generation to the next. That's a key part of it, the part I call 'Communication', which means the information from one generation isn't just totally random compared to the previous generation, but some of the previous generation's information got communicated (by whatever means, it's not important what) to the new generation.

This is how Eukaryotic cell differentiation happens in multi-cellular organisms such as ourselves. Nerve cells come from neural stem cells, not from skin stem cells. That switchboard template of methylation also gets copied during DNA replication.

And there are mistakes/mutations in the copy of the switchboard template, and so it is conceivable that if this mutation rate is at an appropriate level, then it could be enough to fuel evolution of switchboard template patterns, on top of (though entirely dependent on) the underlying DNA genetic patterns.

If the switchboard template patterns last more than a just a few generations, then you have enough stability to enable natural selection of switchboard template patterns.

This may not seem very interesting, and it wouldn't be, if there were only a few sites a which methylation could occur. But with 6 million methylation sites, that's potentially on the order of 6 million switches, which equates to 6 million bits of information, or approx (a bit less than) 1 MB of information.

Still might not seem like much, but then again, each of these switches probably controls the expression of a fairly hefty portion of DNA/genetic information. So, potentially, the epigenetic information can have very significant influence over the organism's environmental fitness.

You might think of it like a cache on a hard-drive (weak metaphor, but go with it). The cache is totally dependent on the information on the hard-drive, but by 'adapting' very quickly to a local environment (the currently running software, I suppose), the cache 'improves' the 'fitness' of the hard-drive (making it work faster and smoother).

And so, whatever underlying genetic mechanisms that handle methylation replication get a boost in fitness if they support epigenetic evolution. Just like hard-drives with caches on them quickly replaced hard-drives with no caches.

I'm not fully convinced, but this article has drastically increased my 'truthiness' feeling about the plausibility of significant levels of epigenetic evolution/adaptation.