The hunt for the genetic roots of common diseases has hit a blank. The genetic variants found so far account in most cases for a small fraction of the genetic risk of the major killers. So where is the missing heritability and why has it not showed up?
A Duke geneticist suggests that the standard method of gene hunting had a theoretical flaw and should proceed on a different basis. The purpose of the $3 billion project to decode the human genome, completed in 2003, was to discover the genetic roots of common diseases like diabetes, cancer and Alzheimer’s. The diseases are called complex, meaning that several mutated genes are probably implicated in each.
A principal theory has been that these variant genes have become common in the population because the diseases strike late in life, after a person has had children. Bad genes would not be eliminated by natural selection at that age, as they would if the diseases struck before the child-bearing years.
So to find disease genes, the thinking went, do not decode the entire genome of every patient, just look at the few sites where genetic variations are common, defined as being present in at least one per cent of the population.
Genomewide association studies
These sites of common variation are called SNPs and biotech companies have developed devices to recognise 500,000 SNPs at a time. The SNP chips made possible genomewide association studies in which the genomes of many patients are compared with those of healthy people to see which SNPs are correlated with the disease.
The SNP chips worked well, the studies were well designed, though enormously expensive, and some 2,000 disease-associated SNPs have been identified by university consortiums in the US and Europe.
But this mountainous labour produced something of a mouse. In each disease, with few exceptions, the SNPs accounted for small percentage of the genetic risk. A second puzzling feature was that many of the disease-linked SNPs did not occur in the DNA that codes for genes, but rather in the so-called junk regions of the genome. Biologists speculated that these SNPs must play an as-yet-undefined role in deranging the regulation of nearby genes.
In an article in PLoS Biology, the Duke geneticist David B Goldstein and his colleagues explain the findings.
They argue that the common disease-common variant idea is largely incorrect: natural selection has in fact done far better than expected in eliminating disease-causing variants from the population. It follows that the major burden of disease is carried by a multitude of rare variants, ones too rare to have been programmed into the SNP chips.
So why have the genomewide association studies linked some SNPs to disease, if in fact it is the rare variants that cause it?
SNPs surrogate markers?
In Goldstein’s view, the SNPs could simply be acting as surrogate markers for the rare variants. Until now, geneticists have assumed a disease-linked SNP was either itself a cause or was a marker for a disease variant nearby. But Goldstein’s team calculated that the rare variants associated with an SNP can occur up to two million units of DNA away from it. This means that the disease-associated SNPs do not necessarily point to anything useful and that it is dangerous to assume the nearest gene is the cause of the disease.
If SNPs are indeed rather indirect markers of disease, that would explain why many have turned up in junk DNA. But why do the SNPs get implicated in the genomewide association studies if in fact it is the rare variants that cause disease? Most of the SNPs are ancient, which is how they got to be common, whereas the disease-causing rare variants are mostly recent, because natural selection is always sweeping them away. After an SNP is created, some of the population has it and the rest continue to carry the standard DNA unit at that site in their genome.
When the rare disease-causing variants build up much later, Goldstein suggests, some will be on stretches of DNA containing the SNP and others on stretches of DNA with the standard unit. Since the allocation is random, more rare variants will be sometimes lie on the DNA with the SNP, and the SNP will appear as statistically associated with the disease even if it is not.
Goldstein calls the association “synthetic”, but it is indirect, so much so as to make many SNPs useless for identifying the genes that cause disease. Geneticists have long been aware of this possibility, but Goldstein’s team has shown that this could happen more often than expected. He has also examined the question in reverse by doing a genomewide association study of sickle cell anemia.
Though the disease is known to be caused by a variant in a single gene, the Duke geneticists found a significant association with 179 SNPs, spread across DNA two and a half million units in length and containing dozens of genes. Most of these SNPs were pointing at the wrong thing.
Genomewide association studies can each cost in the range of $10 million or more. The next step, in his view, is to sequence, or decode, patients’ entire genomes and then to look for likely mutations in the genes.
Finding rare variants useful
Finding even a few of the rare variants that cause disease could point to genes that make suitable targets for drugmakers. The SNPs statistically linked to disease have mostly failed to identify the right genes, but the rare variants may, Goldstein said.
The Icelandic gene-hunting firm deCODE genetics, which emerged from bankruptcy, has long led in detecting SNPs associated with common disease. Kari Stefansson, the company’s founder and research director, agreed that whole genome sequencing would “give us a lot of extremely exciting data.” But he disputed Goldstein’s view that rare variants carried most of the missing heritability. Both deCODE genetics and scientists at the Broad Institute in Cambridge, Mass., have sequenced regions of the genome surrounding SNPs in search of rare variants, but have found very few, Stefansson said.
NYT News Service
