By David N. Leff

¿From months to milliseconds¿ is the time-saving strategy offered to mappers of disease genes by molecular geneticist Gary Peltz, head of genetics at Roche Bioscience in Palo Alto, Calif.

Peltz recites two real-life case histories in which computational analysis economized interminable drudgery and averted cross-breeding hundreds of mice to pinpoint chromosomal regions for susceptibility to asthma and osteoporosis.

¿We were studying a genetic asthma trait with collaborators at Johns Hopkins,¿ Peltz recalled. ¿They were looking at the response in the airways in experimental asthma models of different strains of mice. They identified one strain as an asthmatic responder, the other as a nonasthmatic responder.

¿Using conventional mapping methods at the time,¿ he narrated, ¿we crossed the two strains, generated about 1,200 progeny, then did the genotyping and phenotyping on all 1,200. That took about two years to do, and quite a bit of money. We came up with two genomic regions ¿ one on murine chromosome 2, the other on chromosome 7, which controlled susceptibility to the asthma trait.

¿In comparison,¿ Peltz continued, ¿employing our new in silico computational analysis, we took phenotypic data of the airway response ¿ four parental strains ¿ and ran it through our program. The computer predicted and identified four regions that would control that trait. Two of the loci were the ones we had found in our cross of chromosomes 2 and 7. The other two predicted regions were identified by another group with a different mouse cross. So what would have been four years of work ¿ two on our part and two by others ¿ the computer did in literally milliseconds.¿

His second account concerned mapping a trait for osteoporosis susceptibility. ¿What we did there,¿ he recounted, ¿was work with collaborators at Oregon Health Sciences University in Portland. They had been asking. What are the genetic controls for mineral bone density? And they were using two different strains of mice that had different bone densities. They crossed the pair, analyzed their 1,000 progeny, measured these descendants¿ bone density, then genotyped them. That process alone in their lab took seven months ¿ a not-unusual story.¿

Two True Tales Of Two Disease Traits

¿So we at Roche,¿ Peltz went on, ¿went into our SNP [single nucleotide polymorphism] database, pulled out all the polymorphisms that would distinguish the two strains they used, generated a genome scan, and one lab person ran that scan in less than a week. That was the first time we¿d ever done it. Now that we¿ve got everything set up, it would take less than a day. Most importantly, with that genome scan, and our genotyping methodology, we identified the same four chromosomal regions they had found using the conventional approach. And our computational method picked up a region they missed.¿

These stories are reported by Peltz in today¿s issue of Science, dated June 8, 2001. The article¿s title: ¿In silico [computational analysis] mapping of complex disease-related traits in mice.¿

¿What we¿ve been doing,¿ Peltz told BioWorld Today, ¿is trying to analyze mouse models of common human diseases, to get insight into the pathophysiology. Our ultimate goal is to combine diagnostic ability and therapeutics. So we¿ve developed two tools, described in this paper, that markedly accelerate conventional analysis.

¿The first one,¿ he went on, ¿is a SNP database, which we use to scan the genomes of 15 strains of mice, and characterize where their polymorphisms are. It¿s a fully automated process, computer directed and robot driven. So when we go into a region, the computer tells us what genes are there, designs PCR primers for robot amplification, then prints out programs that analyze the sequences, right as they come off the sequencing machines.

¿The net result of that first tool,¿ Peltz continued, ¿is that what used to take researchers months to perform a genome scan, we can now do in an afternoon. Instead of having to analyze one mouse at a time, we can look at, say 1,000, and pick out 100 high-trait and 100 low-trait animals. Say you¿re studying a disease-related trait, as we did in the bone-density osteoporosis collaboration. We could take out the 100 mice at high density and the 100 at low, and make two pools. So instead of genotyping 200 mice individually, we genotyped two samples.

¿The second, and more important, tool we¿ve developed,¿ Peltz recounted, ¿is a computerized method for performing the genetic analysis. So what used to take months, as we described, can now be done in milliseconds.¿ And what the program does is require information about the disease-related trait measured in various commonly available mice.

¿Where they are found in human populations, the 15 mouse strains had substantial variability. That is, their chromosomal regions harbor genes that express factors conferring susceptibility or resistance for a wide variety of pathologies. These genes can be tracked down by variations in their SNPs.¿

To be sure, mice are not human, but Peltz made the point: ¿The problem with mouse models is not the mice, it¿s the interpretations that people place on the results.¿ He added: ¿There clearly are differences between mouse and man. However, in most of the fundamental traits that we look at, there are similar differences amongst the mice that one sees in the human population. If we can understand the genetic basis for those differences sufficiently in the mouse, the question in an area that still involves some art is how one translates those findings to humans, in a particular context such as a disease association ¿ susceptibility.¿

Roche Extends Pro Bono Access To Researchers

¿Work on generating the Roche mouse SNP database,¿ Peltz noted, is partially funded by NIH¿s National Human Genome Research Institute. ¿We are making that entire database publicly available via the Internet,¿ he said. ¿As of [today], the day Science publishes our paper, researchers can freely access that database. People can query it a number of different ways ¿ such as asking what are the SNPs in a particular region, what are the SNPs that distinguish murine strain A from strain B. All those sorts of questions can be handled in a very user-friendly fashion. We¿re putting our URL address right up on the screen: (¿

No Comments