For more precise drug treatments, ‘squeeze’ the genome: study findsJul. 22, 2021, 10:12 AM
by Bill Snyder
Large-scale studies will be required to identify the complexity of genetic variations that affect how patients respond to a given drug and whether they will have side effects, according to researchers at Vanderbilt University Medical Center.
Such studies are needed both to find new gene variants and to fulfill the promise of precision medicine to improve health, the researchers concluded in a paper published last month in the journal Clinical Pharmacology and Therapeutics.
“At Vanderbilt we’ve gotten really good at using genetic information to help predict drug response,” said the paper’s corresponding author, Sara Van Driest, MD, PhD, associate professor of Pediatrics and Medicine at VUMC.
“The question is whether we’ve squeezed all the information we can out of the genome,” Van Driest said. “If we throw more genes, more variants, into the model, will we do a better job predicting drug response?”
The answer appears to be “yes,” at least for previously assembled datasets of 12 different drug outcomes, or phenotypes, examined in the study. The outcomes involved the use of three cardiovascular drugs, two antibiotics and three immunosuppressive drugs.
Datasets included thousands of individuals who developed acute heart attack or required revascularization (blood vessel bypass or repair) while on cholesterol-lowering statin drugs, and those in whom the use of two common antibiotics and two common immunosuppressive drugs were toxic to the kidney.
Some of the datasets were from the more than 225,000 samples stored in VUMC’s biobank, BioVU. These samples are linked to corresponding electronic health records from which identifying information has been scrubbed.
The researchers found that each of the 12 phenotypes were highly polygenic, meaning that they were associated with numerous genes and genetic variations across the genome, a person’s complete set of genes.
Testing patients for only a few high-impact genetic variations “drastically underestimates” the potential impact of the genome on drug outcomes, they concluded.
Ayesha Muhammad, the MD/PhD student who led the work, said that the results “remind us that even though we have gotten so good at predicting drug response, we have so much more potential to improve.”
“In order to make these improvements, very large cohorts will need to be assembled,” she said. “We also know from prior work that for these predictive models to work well across diverse patient populations, the study cohorts also need to be very diverse.”
“Genomic studies have taught us how we can take giant datasets that contain information that is completely unique from one individual to the next and use them to build predictors and discover ways to make patients’ health better,” Van Driest added.
“The genome is not the only thing that influences how patients respond to drugs,” she cautioned. “I don’t think it is the only component of precision medicine. There will be a lot of things we’ll be able to identify that contribute to drug response” using these techniques.
In the future, predictive models will include non-genetic variables such as socioeconomic factors, drug-drug interactions, drug-environment interactions and perhaps even diet, in addition to genome-wide data.
Van Driest said her group’s research was made possible by foundational programs in pharmacogenetics and pharmacogenomics such as PREDICT and BioVU that were pioneered at VUMC.
Co-author Dan Roden, MD, the Sam L. Clark, MD, PhD, Chair and Senior Vice President for Personalized Medicine at VUMC, is internationally known for his studies of the role that genetic variations play in adverse drug reactions such as drug-induced abnormal heart rhythms.
He is a leader of the Pharmacogenomic Resource for Enhanced Decisions in Care and Treatment (PREDICT), which since 2010 has applied genomic testing to drug prescribing in an effort to avoid adverse reactions.
Roden also is a mentor to Van Driest, Muhammad and co-senior author Jonathan Mosley, MD, PhD, assistant professor of Medicine and Biomedical Informatics, who provided key technical insights in the model analysis.
Other VUMC co-authors were Ida Aka, MPH, Kelly Birdwell, MD, MSCI, and Wei-Qi Wei, MD, PhD.
The study was supported in part by the American Heart Association, Vanderbilt Medical Scientist Training Program and the National Institutes of Health.