Depressing facts not to get depressed over: zero genes for depression?

Depression seems to run in families, but despite numerous attempts, few causal genes have been identified and even fewer have been replicated.  Why?

Twin and family studies have estimated heritability, or the proportion of the cause of depression that is genetic, to be 31-42%, according to a newly published report in Biological Psychiatry.  This is a report of a meta-analysis that pooled results from 17 genomewide association studies (GWAS), including a total of 34,549 people of European origin with depression somewhere along a continuum defined by the answers to 20 questions as to extent of depressed mood, feelings of guilt and worthlessness, feelings of helplessness and hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance.  This study differs from most others in that it used symptoms rather than diagnosis.  

They also attempted to replicate their findings with 5 studies that used a different assessment of depressive symptoms, the requirement these days for validating a study. Finally, they did a combined meta-analysis of a number of discovery and replication studies, for a total of 51,258 individuals.

 Summed up bluntly by Science News:

A massive effort to uncover genes involved in depression has largely failed. By combing through the DNA of 34,549 volunteers, an international team of 86 scientists hoped to uncover genetic influences that affect a person’s vulnerability to depression. But the analysis turned up nothing.

Well, not exactly nothing. The authors report one hit when they pooled all the studies, but it did not replicate any other studies, and it was in a genomic region that included no genes. That's not necessarily meaningless, but it is difficult to follow up.

 The authors wonder whether using the depression scale might explain this lack of hits. Were their cases so heterogeneous that this prevented them from finding a major effect? If they'd included only people who scored high on the scale, perhaps this would have narrowed down possible causal genes, or eliminated cases of depression that might not have a genetic basis. The authors wrote:

The approach of studying depression on a continuum has the advantage that not only information on extremes is used but that all available information is exploited. Van der Sluis et al. showed that if the phenotypic variation among cases, as well as the variation among control subjects, is used, this greatly increases the power to detect genetic variants. 

But, given that no other study has replicably found genes with major effect, it is unlikely that this explains the lack of significant findings. Their replication studies may have added a different phenotype to the mix, too, and therefore additional genetic heterogeneity. The authors also point out that gene-gene interaction or gene-environment interactions might explain depression, and hinder GWAS.

 Of course, the authors (naturally, and with total scientific disinterestedness) say that the way to do this right is to do a larger study of, say, 50,000 cases. But of course this won't eliminate any of the issues, and will only increase the heterogeneity problem.

Naive expectations
Showing that this isn't just evidence of sloppy workmanship is the fact that similar results have been found for most other psychiatric or neurological disorders, with few exceptions, and that even 'mechanical' traits like the skeleton, metabolism, and so on, that are easier to define and measure show the same level of complexity. Even in yeast.

Is it fair to point out that we have known for a mere 94 years that it is absolutely consistent that traits without major genetic contributions can cluster in families, so that substantial heritability is not an indicator that mapping will find such genes. While families with many cases in many generations and collateral relatives (like cousins) do raise hopes of such effects at least in those families, it was long ago also shown that polygenic causation (which is what we have here, if the evidence is to be believed) can mimic Mendelian transmission families.

So this study, like so many costly others, tells us what we knew. Indeed, we already knew we knew that before this (and so many other) study was done.

When nothing is something--unless it's not something you want!
We're in the age in which we boast about our various name-dropping technologies that we used in our latest elephantine study. We make sure our listeners know that we've used 'massively parallel' or 'Next-generation' sequencing to get data from massive numbers of people. The idea is that signal-to-noise ratios are well-behaved, so that increasing sample sizes will find increasingly weaker signals.

There is no reason to believe this, when it comes to genetics, and indeed there are good reasons to think the contrary. But that doesn't slow down the current modus operandi: generating massively incremental data.

Actually, on the other hand, the 'nothing' that is being found is nothing to be depressed about! Instead, it's convincing evidence for complexity--that is, that many minor causes interact and combine to generate results, so that each case is different in detail, both in terms of the trait itself and the genotype that contributes to it. It's the nothing we don't want because what we want is something we can cash in on: build a career as the discoverer of the Big Fact, make and profit from a patent, make a pharmaceutical bonanza.

Instead, it's the mouse that roared. We should be taking credit for showing how life really seems to be, which all these GWAS have in fact shown, but few are willing to accept it. Of course, we needn't do that thousands of times, except that that's what we're doing because of the paucity of better ideas. But that's how humans, at least humans in our kind of culture, seem to operate.