A computer may be worth 1000 mice if a new genetic mapping technique pays off. The approach could markedly speed up the first step in identifying genes associated with diseases, making the process cheaper and more efficient.
Scientists often search for disease genes in inbred strains of laboratory mice, which have less "genetic noise" than human populations. By breeding hundreds or thousands of mice, they can look for genetic markers associated with traits like obesity or high cholesterol. The patterns of association can identify regions of the mouse genome, called quantitative trait loci (QTL), that likely contain genes that contribute to the trait. This is a laborious process, taking years just to get a rough idea of where a gene resides.
Now, a team of scientists has come up with a way to speed things up. Gary Peltz of Roche Bioscience in Palo Alto, California, along with colleagues at Roche, Stanford University, and Oregon Health Sciences University in Portland compiled a database of more than 3000 genetic markers called single nucleotide polymorphisms (SNPs) for 15 inbred mouse strains. Then the team created an algorithm that lets them query the SNP database to identify QTLs almost instantly.
Here's how: A researcher first looks up or measures how a particular trait, like cholesterol levels, varies among a number of the strains. The algorithm then compares pairs of strains, looking for SNP patterns that are shared among strains with similar phenotypes, but not shared among strains with different phenotypes; the more strains that can be compared, the better the prediction.
To test the algorithm, the researchers fed published phenotypic data for 10 traits--bone density and tendency to consume alcohol, among others--into the computer. Then they checked the computer-predicted loci against published QTLs mapped through the conventional process of mouse breeding. The predictions matched the actual locations 75% of the time.
That's not proof positive yet, cautions Dean Shepherd of the University of California, San Francisco. "To prove what the method is really worth, we'll have to actually find some specific mutations that explain the differences in phenotype," he says. But Shepherd is optimistic, saying, "It's extremely likely that in the near future this will really have a significant payoff."