Willer et al. (2008). Six new loci associated with body mass index highlight a neuronal influence on body weight regulation Nature Genetics DOI: 10.1038/ng.287
Thorleifsson et al. (2008). Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity Nature Genetics DOI: 10.1038/ng.274
There are two massive studies now online in Nature Genetics looking at the genetic architecture of body mass index (BMI). Body mass index is a widely-used measure of body fat levels, calculated by dividing a subject’s weight (in kg) by the sqare of their height (in metres). Individuals with a BMI above 30 are classified as obese.
BMI is substantially influenced by genetics, with a heritability of 40-70%*, but the precise genetic regions responsible for its variation have proved remarkably tough to uncover – despite large, genome-wide analyses of common genetic variation, only two regions (near the FTO and MC4R genes) have shown convincingly replicated associations with BMI.
The two latest studies tackle BMI with sheer brute force of numbers: each of them report genome-wide association analyses (combining new data with meta-analysis of published data) of over 30,000 individuals, with follow-up studies in as many as 60,000 samples. That power allows both groups to identify a scattering of novel regions with an effect on BMI – 10 new regions in total, with four regions identified in both studies.
So, how do these new variants perform in terms of allowing us to predict BMI based on genetic data alone? Here’s the breakdown from one of the papers:
In our stage 2 samples, the six newly discovered loci together account for 0.40% of the variance of BMI, and in conjunction with the known associations at FTO and MC4R account for 0.84% of the variance.
In other words, all of the genetic variants identified so far, despite genome-wide studies of unprecedented scale, explain less than 1% of the overall variation in BMI.
The same paper also reports using their genetic data to come up with a “genotype score”, which can be seen as a kind of genetic risk profile for individuals. Even combining all of the available genetic variants the predictive power of this score is pretty meagre:
…the 1.2% (n = 178) of the sample with 13 or more ‘standardized’ BMI-increasing alleles across these eight loci is on average 1.46 kg/m2 (equivalent to 3.7-4.7 kg for an adult 160-180 cm in height) heavier than the 1.4% (n = 205) of the sample with less than or equal to3 standardized BMI-increasing alleles, and 0.59 kg/m2 (1.5-1.9 kg for an adult 160-180 cm in height) heavier than the average individual in our study.
In other words, the 1.2% of individuals with the “fattest” genes were only 2 kg heavier on average than individuals with “average” genes. That really is pretty minimal predictive information.
It’s likely that further increases in sample size will yield more common variants with small effect sizes, but we’re now well into the zone of dimishing returns – given the size of these studies, most of the common variants explaining more than 0.1% of the variance in BMI are likely to have already been found. However, fine-mapping of the associated regions (using a more high-resolution set of genetic markers to zoom in on the causal variants) may identify better predictive markers, and further studies looking at lower-frequency variants will likely start to chip away at the remaining unexplained variance. These studies have already identified some variants that will probably prove to be genuinely causal, such as an large insertion-deletion polymorphism close to the NEGR1 gene – one of the first convincing associations between a so-called structural variant and human physical variation.
In any case, the most important information to emerge so far from genome-wide association studies has not been strong predictive markers, but insight into the sub-cellular pathways underlying human diseases and traits – for instance, one of the intriguing results from these studies is that the genes most closely associated with BMI-linked regions are not enriched for involvement in energy metabolism, but rather for expression in the central nervous system. Uncovering the causal variants underlying these signals (through a combination of genetic and functional work) will provide unbiased insight into the molecular mechanisms underlying human obesity.
* Update: Razib covers the same study, and makes an important point about the interpretation of the term “heritability”:
Jake Young & I have talked about the problems with the public interpretation of a quantitative genetic number which refers to heritability of BMI on the order of 50%. Most people interpret this to mean that “obesity is 50% genetic,” but this really means that half of the variation in BMI within a population is due to variation in genes. The reason people are way fatter today than they were a few generations ago likely has little do with genetic differences, and much to do with environmental differences. In fact, the heritability, or proportion of population level variation in BMI due to genetic differences might well be increasing because of the change in environment!