Growing Pains for Field of Epigenetics

Demethylating agents, HDAC inhibitors, and their like can induce powerful effects on malignancies via (purportedly) primarily epigenetic mechanisms. I&#x27;ve seen patients with astronomically high leukemic counts return to normal blood counts on a few weeks of romidepsin, for instance.<p>But when groups use epigenetics to study poverty related stress, risk of depression, etc, there is a very different political structure than when comparing drugs to see what kills cancer cells in a dish. The trend seems to be to publish borderline findings with a nature vs nurture argument to explain differences as environmental, not genetic, and call for action as well as <i>more funding</i> to expand the research and find ways to environmentally or behaviorally prevent the problem. Optimistically, it&#x27;s trying to solve problems. The issue is that important findings will be mixed in with a lot of questionable results that sound appealing to liberal academic journal editors and get a free pass at publication in top journals, a process which feeds back into the SJW gravy train of getting more academic grants to do more of the same. And with sciency techniques and big data approaches, what could be more fashionable? It really does a disservice to the subset of epigenetic research which is well conducted and reproducible. If the epigenetics bubble pops a bit, good. Other fields could use the attention.

As I see it pretty much from inside, this is not a general problem but just the evolution of science, as we are able to determine more and more things which lead to better and better conclusions. The problem in what is cause and effect comes from the sheer amount of data you create (for a current study we have whole genome methlaytion data for 19 cell populations in mice, all in triplicates and at medium coverage), which is obviously prone to many false positives. And as we are now approaching single cell level (which will dramatically improve results), this number is only going up. And of course it is hard to check all these positives extensively. But yeah, this is sciece and we are only getting better, so no pain but just an opportunity. A problem is rather the amount of data we create (we are speaking about Petabytes) that have to be stored and made accessible for decades so we can later recheck our conclusions. Nobody wants to pay for that

I am working with some epigenetics data right now (chip-seq on histone modifications). The main problem I see is that there doesn&#x27;t seem to be any mature framework for analyzing the data. There are a bunch of disparate methods for different aspect of the data analysis that all seem to give different results.<p>From the data generation perspective, there are so many possible sources of biases from preparing the biological sample to sequencing of the sample, it can be difficult to control for all these variables in the down-stream analysis.