A computational genetic analysis method developed by scientists at Roche has inched a bit closer toward industrial feasibility after the researchers this month published a pair of papers validating the approach.
The two papers — one in Anesthesiology on disease propensity and a second in Nature Biotechnology on drug metabolism — served as proof of concept of the approach, which uses mouse haplotype information to hone in on genetic factors that contribute to strain-specific phenotype differences.
Gary Peltz, head of genetics and genomics at Roche and co-author on both papers, stressed that the computational method is still at an early stage. Nevertheless, he told BioInform, the approach could have "significant implications when it's more fully developed and utilized. We hope that in the future, it will become one of a number of new genomic tools that will help us better understand our drugs, as well as our diseases."
In the Anesthesiology paper, Peltz and colleagues use the method to identify gene variants associated with the likelihood of developing opioid-induced hyperalgesia, an exaggerated sense of pain associated with morphine addiction.
In the Nature Biotechnology paper, the Roche scientists use the approach to predict genetic variants responsible for individual metabolic response to the blood-thinner warfarin.
"We hope that in the future, it will become one of a number of new genomic tools that will help us better understand our drugs, as well as our diseases."
Both studies were performed on inbred mouse strains. The method builds on a method that the Roche team published in 2004 called HapMapper, which relies on data in the company's freely available mouse SNP database BioInform 10-25-04.
The approach "computationally predicts causal genetic factors by identifying genomic regions where the pattern of genetic variation correlates with the distribution of trait values among the inbred strains analyzed," the authors wrote in the Nature Biotechnology paper. The technique uses ANOVA-based modeling to correlate between trait values — whether opioid-induced hyperalgesia or warfarin metabolism — and the strain groupings for each haplotye block.
Peltz said that the approach "markedly accelerates the rate at which you can perform genetic analysis in mice." Since it was published in 2004, the Roche team has been trying to apply it in two primary directions, he said: "understanding diseases and understanding drugs."
The recent papers serve as initial validation for both application areas, he said. While the team's 2004 paper highlighted the ability of the method to identify disease-related genes, the Nature Biotechnology paper marks the first use of the method in a pharmacogenomics application.
Human metabolic response to warfarin is fairly well understood, making it an ideal subject for a proof-of-of concept study on drug metabolism, Peltz said. Genetic variants of the CYP 2C9 enzyme are associated with different rates of metabolism of the drug, which can have drastically different effects in different people.
Warfarin is "the most common cause of drug-induced problems in the clinic," he said. The drug has a "very narrow therapeutic index. If you give somebody too much, they can bleed, if you give them too little, they can clot, so it's a high-risk situation."
In the mouse study, the Roche scientists correlated genetic variations in CYP 2C29 — a "close cousin" of CYP 2C9 — with strain-specific differences in warfarin response.
"It was a proof of concept giving a result that was in line with an expectation," Peltz said.
The complexity of the drug metabolism process did present one challenge, according to the authors. "Warfarin metabolism is a very complex process — its biotransformation to over nine different metabolites in rats and humans is mediated by many different enzymes. Each individual genetic difference could be responsible for only a small portion of the inter-strain differences in warfarin pharmacokinetics," they wrote.
However, in order to work effectively, the Roche computational approach requires a single genetic change to have a large impact on a phenotypic trait. "Simulations have shown that an individual genetic factor must be responsible for at least 40 percent of the phenotypic variation for it to be identified by haplotype-based computational genetic analysis when 13 to 15 inbred strains are analyzed," the authors noted.
As a result, they developed an experimental strategy based in radiolabeling to reduce the complexity of the metabolic process. "We used radiolabeled drug and then followed the individual metabolites," Peltz said. "So what you're doing is taking a process that's multidimensional and following a specific component in it, and that's what made it genetically mappable."
Going forward, Peltz said that the method should prove valuable for predicting genetic factors that play a role in systems that are not as well understood as metabolic response to warfarin.
"For many of the drugs that are commonly used, we know relatively little about the genetic variables affecting their metabolism, or more commonly their mechanism of action or their toxicity," Peltz said. "So what we hope to use this experimental mouse system to do is to be able to identify in a fairly quick way what are the genetic variables that affect either metabolism, mechanism of action, or toxicity."
Peltz said that the Roche team will continue to improve the method to make it more applicable to industrial use. The next step, he said, is to move the system toward an in vitro platform. "Right now, we administer drugs, draw, and analyze in the plasma," he said. By moving the approach to a microtiter well format, "the rate of data generation just exponentially increases," he said.
In addition, the Roche scientists will apply the approach to "more complicated problems," such as toxicology. "We're taking a look at a couple of other drugs that cause liver toxicities and seeing if we can understand the genetic factors affecting susceptibility or resistance to drug-induced liver diseases," Peltz said.
"Drug-induced toxicity is often caused by intermediate metabolites, rather than by the parent drug itself," the Roche authors noted in the Nature Biotechnology paper. Therefore, "correlation of the strain-specific pattern of drug-induced toxicity with the production of certain metabolites, as well as with a pattern of genetic variation within a gene, can be used to identify a genetic susceptibility factor for a toxic response."
Peltz noted that while the approach itself is not applicable to humans, the data gleaned from mouse studies should serve as a useful starting point in identifying human genes of interest in a range of applications.
"If you know where to look, you can get there faster," he said.
— Bernadette Toner ([email protected])