Linking out to a book review on sociogenomics

The journal Nature has a review of the new book, Social by Nature: The Promise and Peril of Sociogenomics, by Catherine Bliss. The review is written by Nathaniel Comfort: “Nature still battles nurture in the haunting world of social genomics”.

After reading the review I plan to read the book and bring some of this topic into my genetics course this semester.

Comfort’s review presents Bliss’ book as a critical survey of work in social genomics. I can’t assess from Comfort’s review whether the book is really accurate about the genomics research that it describes, and that’s my main interest.

That’s because a historian of science writing about a sociologist uses a hecka lot of jargon.

For instance, the “genomic gaze”:

What the historian Andrew Hogan has called the “genomic gaze” isn’t the fault of individual bad-guy researchers: it’s structural. Bliss is careful to acknowledge the good, even noble intentions of many of the scientists she spoke to (as a sociologist, she keeps the names of her ‘informants’ confidential). But she finds that the funding and publicity mechanisms integral to biology drive it towards genes-first explanations. The stakes are high: finding an SNP associated with a risk increase from 0.01% to 0.03% (a threefold rise) for a disease such as breast cancer could make a career. “While researchers do not intend to lift the focus off of the environment,” Bliss writes, “they are forced to recast social phenomena as ‘evolutionary phenotypes’ so that they can make scientific claims” that sound relevant to biomedical funders.

It seems like sociologists always disappoint me when they talk about the motivations of geneticists. “Finding a SNP” is hardly likely to “make a career” these days.

To be sure, like other areas of science, genetics has a publication bias toward positive findings. But during the last fifteen years genomics has raised the standard of evidence necessary to publish positive associations. A good discussion of this subject would describe the candidate gene approach, still heavily used in some areas of behavior genetics. The candidate gene approach has been strongly criticized within genomics, as genome-wide association studies demand vastly larger sample sizes and result in stronger evidence of association.

This is relevant because today many of the most interesting findings in human behavior genetics are coming from large GWAS-scale samples of tens or hundreds of thousands of research subjects.

The paragraph I’ve selected from Comfort’s review uses as its example a fictitious example chosen to illustrate a very small effect size (“a risk increase from 0.01% to 0.03%”). But “breast cancer” is a pretty poor fit to this fiction. The lifetime incidence of breast cancer in women in the U.S. today is more than 12 percent, three orders of magnitude higher!

Numbers matter. A risk of 0.02% is two cases in every ten thousand people. A study of a candidate gene with 80% power to find a significant difference between 0.01% and 0.03% would have more than 60,000 cases and controls. Geneticists who study rare disorders that actually have such low incidences make extensive use of pedigree data to find genetic regions of interest. Finding a genetic influence so small out of a genome-wide association study is not statistically credible.

In other words, if it seems ridiculous to focus on such a small difference, that’s because the contrived example is nonsensical.

Comfort’s general point has some truth to it. Human genomics has led to the discovery of many genetic variations that have small effects on traits. The “one gene one trait” model, such a wide misconception among the public, is wrong. Meanwhile, the best summaries of genetic associations today still leave a large fraction of heritable variation unexplained.

But those issues have been advanced by professional geneticists, not sociologists. During the last year, one of the most important theoretical questions in quantitative genomics has been whether such small effect size mutations actually matter to anything. They may simply be evolutionary noise. And still, the variation in many human traits is the sum of hundreds of such small genetic effects. Those genetic effects make people different from each other. Some of them seem to have been targets of natural selection in the last couple of generations.

I’ll be reading Bliss’ book carefully to see how she describes the results of behavior genomics. Also I’m curious to see how she describes the origin of the term, “sociogenomics”. It’s not one I’d ever heard before seeing this review.

UPDATE (2018-01-17): As I was writing this, I saw a tweetstream by Jeremy Freese, a social scientist who is quoted in the book. He notes numerous apparent errors of fact in the text, from misspelled names and incorrect dates, to paraphrases of quotes that differ from their sources.

I have no further information but it is a point of concern.