Why molecular genetics will not replace quantiative genetics

Today, at our lab-meeting, I briefly mentioned an interesting bloggpost by Juha Merilä at his research group's blogg, that I would like to discuss a bit. The reason is that it is not only interesting from the viewopoint of quantitative genetics vs. molecular genetics per se, but also of its wider implications in terms of the value of reductionism in science vs. emergent properties of biological systems. What do I mean by this?

Emergent properties of systems simply means that "the whole is more than the sum of the parts". This is nothing metaphysical, but it means that one might not fully understand a system simply by de-composing it in to separate parts, because the parts often interact with each other. One actually needs to understand how the parts work together, and it is here where I see the value of quantitative genetics. Quantitative genetics is focussed on the "units" (=traits) that we as evolutionary biologists are most interested in and which we would like to understand how they evolve. It focuses on these higher-level units, while at the same time ignoring the details (=the particular single genes governing the traits). This might be perceived as a weakness of quantiative genetics, but others would argue that it is a strength. I will not dwell in to this here.

Although none of us would deny that the traits are governed by many different genes, as well as the environment of course, we do still not fully understand the genetic basis of the vast majority of traits, in spite of the explosion of molecular information and the presence of fully-sequenced genomes. And if you ask me, I am not sure we will necessarily get the answers to these fundamental questions in biology simply by sequencing even more, or by hoping that new molecular techniques will solve all the remaining questions. In short: the whole will always be more than the sum of the parts, and quantitative genetics is still live and kicking. At least in some circles.

Back to Juha's bloggpost. An interesting "paradox" is that quantitative genetic studies of human height indicate that heritability of height is very high: between 80 and 90 %. In other words, height is a highly heritable trait that can rapidly evolve by natural or sexual selection. One would therefore think that height would be an excellent trait to focus on if we wish to find candidate genes or do genome-wide association studies to identify the precise genes. Such a study has indeed been performed, but as the blogg "Genetic Future" rather brutally points out, these studies have been "a resounding disappointment".

Some more specific examples of these disappointing results:

The first genome-wide association for height, published in September last year, examined 4,921 individuals and found a single significant and replicable variant associated with height (in the HMGA2 gene) that explained a meagre 0.3% of the variation in the general population. The second such scan, published in January this year, examined 6,669 individuals. This study confirmed the HMGA2 finding and identified one further significant signal, this time near the GDF5 and UQCC genes, which again explained less than 0.5% of the total variation. I saw an abstract at the American Society of Human Genetics meeting last year (as yet unpublished, as far as I can tell) in which a genome scan was reported for 10,737 individuals, which pulled out a total of 8 associated variants which together explain just 3% of the variance in height.

To summarise these results: to date genome-wide scans for height, even extremely well-powered ones with more than 10,000 participants, have identified variants responsible for less than 5% of the variation in this trait - despite height being a trait that is largely genetically determined and varies substantially between humans. What's going on?

Well, I certainly do not know the answer, and neither does Juha or the blogger "Genetic Future". But we can safely conclude that a major fraction of the genetic variation in height is "missing" and not picked up in these genome-wide association studies and by the molecular markers. Hopefully, future studies can improve resolution and increase these depressingly low percentages somewhat. However, these low percentages should be somewhat worrying, at least for those who thought that it would be easy to study the molecular basis of this trait. Especially given that we already know that there is ample genetic variation for human height from the quantitative genetic studies.

It seems to me that either the heritability estimates have been severely inflated (e. g. by maternal effects), or the molecular markers do not fully capture all the genetic variation in human height. Although it is definitely possible that heritability is somewhat overestimated, I doubt that heritability will only be a few percent, as the molecular marker studies indicate. Most morphological traits have heritabilities much higher than a few percent, so it seems a safe conclusion that the molecular markers miss a large amount of the genetic variation that is actually present. Therefore, it seems as if the association studies are missing the big picture. The results from these genome-wide association studies must in any case be a big dissappointment, given the large research resources that have been put in to them and the HUGE sample sizes in these studies.

The EGRU-blogg cites a recent update about this work in Nature, and (musingly?) concludes:

"Read the thought provoking ‘news feature’ from Nature by Brend Maher about the statue puzzle. It might even give some (quantum of?) solace for those sticking with their parent-offspring regressions and tedious experiments while the others in their spotless labcoats are drowning in their high throughput data."

Juha Merilä also comments on another interesting bloggpost with the title "Do we still need quantiative genetics in the 21st Century":

"It seems that people were a bit optimistic about the utility of the new technologies in solving old problems. This has also lead to situation where students have jumped into the fancy wagon of genomics, and left quantitative genetics with some interesting (and unfortunate) consequences of which you can read from here:

http://www3.interscience.wiley.com/cgi-bin/fulltext/119408495/HTMLSTART

Actually, all the 2008 volumes of the Journal of Animal Breeding and genetics feature nice Editorials each which touch some general issues interesting also for Evolutionary Biologists. Go and have a look:

http://www3.interscience.wiley.com/journal/118521919/issueyear?year=2008"

I, myself, is like Juha a big fan of quantitative genetics, and I do of course agree with him and his colleagues that this is a useful tool that is unlikely to be replaced by molecular genetics, neither in the short- nor in the long-term. As a matter of fact, molecular genetics might have its greatest utility as a complement to quantitative genetic approaches, e. g. in the combination of QTL-approaches and estimations of quantitative genetic parameters such as to understand the "anatomy" of genetic correlations. Are genetic correlations mainly a result of linkage disequilibrium or pleiotropy, is one such interesting question that can be adressed by the combination of quantitative genetics and QTL-approaches.

And it might of course always be good to keep in mind that one should be critical of scientific bandwagons and premature claims that new techniques will solve all problems. Finally, do not forget the emergent properties and that the whole is always more than the sum of the parts. Reductionism can be taken too far, particularly in biology.