Another uninterpretable epigenetics study

If you looked at your Twitter feed on Sunday you likely saw a lot of buzz about a new study that found that “Holocaust survivors trauma is passed on to children’s genes”. Many people have already taken time to blog about the issues with this study, but I wanted to ensure that the message was passed on to the molecular ecology community because I think that it is relevant to the community. So here is a quick (and by no means comprehensive) list of why you should be skeptical about the study (and, importantly, why the authors should not have been able to draw the conclusions that they drew):

1) A tiny samples size: they had n=32 Holocaust survivor parents and n=8 controls. It seems like it would have been easy to add some more controls – perhaps even matching the number of Holocaust survivors and their offspring (clearly the limiting subjects).

2) No correction for multiple hypothesis testing: They conducted at least 6 tests using the same data and obtained uncorrected “significant” p-values of just barely below 0.05.

3) No control for the tissue composition: Peripheral blood is a heterogenous tissue and we all know that DNA methylation is highly tissue-specific. So it is possible that they could have just spent way more money than they needed to calculate the proportion of different blood cell types in their samples.

4) No control for genetic background: methylation can be altered by local genetic variation. While the authors looked at one very specific “risk allele” in their gene, they did not test for the effect of other variants. In fact, the strong correlation between F0 and F1 methylation patterns suggest that methylation in this region has a some genetic component that probably isn’t captured by genotyping just one site.

5) Figures 3 and 4 (see images below): Do you see the outliers in both of these figures? The correlations in both genotypes in Fig. 3 seem to be driven by 1 or 2 outliers. Remove those points and then redraw the lines… see? Also, the significant difference on the y-axis in Fig. 4 looks like it passes a p-value of 0.05 because of that blue circle all the way up at ~70% methylation. I’m focusing on the y-axis here because their main finding is that there is a difference in the means of the two groups on the y-axis. It would be nice to have more data here, especially in the control group, to ensure that the findings are robust (the jury is still out).

Fig. 3 in the Yehuda et al (2015) paper.

Fig. 4 in the Yehuda et al (2015) paper

Despite all of these caveats the authors still drew the conclusion that:

To our knowledge, these results provide the first demonstration of transmission of pre-conception stress effects resulting in epigenetic changes in both exposed parents and their offspring in adult humans.

So, save yourself the time and money and don’t bother pyrosequencing [insert locus of choice here] to see if [insert early life and/or environmental variable here] influences DNA methylation at just one site. It’s going to be hard to interpret.


About Noah Snyder-Mackler

I'm a postdoctoral fellow in the department of Evolutionary Anthropology at Duke University. Broadly, I study non-human primate genetics and genomics. More specifically, I'm interested in the interaction between behavior, genotype, and gene expression in response to social stress.
This entry was posted in Uncategorized. Bookmark the permalink.
  • ilovepigenetics

    There’s even more wrong with this paper (and my background is in epigenetics).

    1. In order for this to be transgenerational, the effects in this study must be observed in the F2 generation, as the F1 gametes are exposed.

    2. The DNA methylation changes are different in the F0 and F1 generation. If they were the same changes, that would be more interesting. The fact that they are different (and the N in the controls is TOO SMALL), suggests that this is noise.

    3. The DNA was extracted from the blood of the F1 generation during the study, but from archived samples in the F0 generation. This in itself could account for some of the differences between the generations, and the small N in the controls may not allow the authors to pick that up.

    4. Conducting this type of study without breaking out males and females presents a serious flaw. The authors state there are no differences between males and females, but
    the number of controls is too small to state this. There are only 2 male
    Holocaust survivors and 1 male progeny.

    There are more problems. I REALLY wanted to read a great epigenetics paper… This one isn’t it.

    • Noah Snyder-Mackler

      Thanks for adding to the list! #1 and #2 are especially important.