Researchers at the Institute for Systems Biology have developed a method that can predict how combinations of gene mutations affect the way yeast cells respond to environmental stress.
Their work, which appears in the March 27 issue of Molecular Systems Biology, may help investigators understand at an earlier stage in the drug-development process the complex range of effects that perturbations have on gene expression. This knowledge may enable them to model the perturbations’ intended and unintended effects on the target gene.
“Our methods are designed for application to any system in which multiple interacting genes are linked to phenotypes,” said Greg Carter, a research scientist at ISB who helped develop the method.
Tim Galitski, an associate professor at ISB, added that in order to predict how perturbing a biological system or cells in a specific way will influence a phenotype, researchers must consider not only the direct effects of the individual perturbations on genes and gene expression, but also the effects they have on other perturbations and gene variants in a genome.
“One of the biggest problems in drug discovery is there are a lot of compounds that make it pretty far along in the process, but then do not receive [US Food and Drug Administration] approval,” Galitski said. Failing to consider how these different perturbations interact plays a role in these failures.
Galitski went on to say the drugs that make it to a certain point in the development process are probably efficacious, but may have side effects that would hurt their chances at being approved. He said the kinds of studies that he and his colleagues performed may help to predict these adverse events.
“These kinds of approaches might help to rescue some value from many drug candidates that have fallen by the wayside,” said Galitski. “A drug is, in the end, a genetic perturbation, and can be modeled as such using the method that Greg developed.”
He said that this technique may be used to determine if combinations of drugs are safe, even if an individual drug is not.
Carter said his team used the filamentous growth response of budding yeast as a model system. He explained that in response to environmental cues, yeast cells switch from their round, single-cell growth form to a pathogen-like form that is both invasive and adhesive to its substrate.
The investigators collected microarray data of multiple genetic perturbations under these conditions, Carter said. Viewing each gene as a quantitative phenotype, they subjected the expression data to mathematical decomposition.
Cater said that the genetically direct effects from regulator genes on many differentially expressed genes were parsed out from the indirect effects involving interactions between regulator genes.
They next integrated molecular interaction data from their decomposition results to construct regulatory network models, said Carter. He said that when the investigators tested a set of predictions with additional microarray experiments, they found that their model provided more accurate predictions compared to a similar model that did not incorporate genetic interactions.
“Our methods are designed for application to any system in which multiple interacting genes are linked to phenotypes.”
The researchers were then able to identify an expression pattern that was strongly correlated with measurements of the filamentous growth phenotype.
“Any gene was a potential candidate for analysis in our study,” said Galitski. “In practice, we observed that approximately 1,800 out of about 6,000 genes in the yeast genome changed significantly during the study with the perturbations that we used.”
An Eye to the Future
Carter said that the ISB team is currently following up on a couple of areas. He said that in addition to deletion perturbations, the researchers would like to look at the effect of different dosages and the overactivity of genes. As he explained, “when you introduce small molecules or drugs, dosage is going to matter, so we want to learn how to model these effects as well.”
The investigators would also like to extend these techniques to look at more natural variant strains, Carter said.
“For example, we could take crossbred strains of yeast and look at the progeny,” he said. “We would know that we have a limited, randomly mixed gene pool. We want to further develop the technique so we would not need to know exactly what the engineered perturbations are going in. We could then begin to dissect some of the direct and indirect effects in a more naturally variant genetic population.”
Galitski said that the team wants to study a situation that is more representative of the human population.
According to Carter, the group will probably be publishing follow-up work within the next four to six months. That data will likely look at the effects of dosage variations, he said.