A new approach to personalized medicine examines the activity of a whole group of genes to predict a patient’s sensitivity to drugs. The results of the study, published by researchers at MIT in the journal Genes and Development last week, may help better match cancer patients with the most effective chemotherapeutic drugs while also keeping side effects on healthy cells to a minimum. The technique could also help determine an individual’s sensitivity to environmental toxins, such as those contained in cigarette smoke.
Two approaches that are commonly used to determine how a patient’s tumor will react to chemotherapy include laboratory tests of the tumor cells’ responsiveness to the drug, and genetic screenings for mutations or activity levels of genes that are known to influence drug sensitivity.
But Leona Samson, director of MIT’s Center for Environmental Health Sciences and senior author of the current study, believes that her new technique may offer a quicker and more accurate diagnostic approach.
“Directly looking at the biological response of cells to drugs is a delicate and time-consuming process,” she says of the first approach. The second method–looking for changes in specific gene, works well for certain cancers, she says, such as glioblastoma–the brain tumor that Ted Kennedy is battling–but the accuracy is limited to around 60 percent because other genes will also affect the tumor’s sensitivity.
In glioblastoma, specific markers that reveal the expression of MGMT, a DNA-repair enzyme that makes tumors resistant to the DNA-damaging drug MNNG (N-methyl-N’-nitro-N nitrosoguanidine). MNNG can damage and mutate healthy cells by introducing errors into their DNA. Drugs like MNNG are used in chemotherapy because it’s especially toxic to cancer cells, which often have reduced DNA-repair abilities, but the compound and its close relatives are also found in tobacco smoke, exhaust fumes, and some foods.
To overcome this limitation, Samson and her colleagues used microarray analysis to discover if looking at many genes in parallel could improve the prediction. They took 24 blood-derived cell lines from healthy and unrelated people with diverse genetic backgrounds and tested how well they grew when exposed to MNNG.
Next, the researchers analyzed the activity of 20,000 genes in each cell line by measuring the number of gene transcripts on DNA microarray chips. Finally, they chose the four most sensitive and the four least sensitive cell lines and used them to train a computer algorithm to identify which of the 20,000 genes are most relevant to drug sensitivity. These genes showed different expression levels in the sensitive versus resistant cell lines, and their activity either gradually increased or decreased with the sensitivity of the cells.