Personalized Medicine Starts Predicting Personalized Side Effects

Personalized medicine is usually about personalized benefits. For example, a drug’s potential to benefit a particular patient can be determined on the basis of that patient’s unique genetic make-up, as indicated, for example, by a genomic or metabolomic profile. But what about a drug’s capacity for side effects? If personalized medicine can predict, on a patient-by-patient basis, a drug’s upside, it can, presumably, also predict a drug’s downside.

This notion was put to the test by researchers at the University of California, San Diego. They conducted a proof-of-concept study to demonstrate the feasibility of predicting a drug’s side effects on an individualized basis. Specifically, they developed a predictive model that relied on measurements taken from patients’ blood.

The study was based on genomic and metabolomics data obtained from blood samples of 24 individuals. Researchers used these data to build a personalized, predictive model for each individual. Researchers then used these predictive models to understand—at the metabolic level—why some individuals experienced side effects to ribavirin, a drug used to treat hepatitis C, while other individuals did not.

Details of the study appeared October 28 in the journal Cell Systems, in an article entitled, “Personalized Whole-Cell Kinetic Models of Metabolism for Discovery in Genomics and Pharmacodynamics.”

“[We] constructed multi-omic, data-driven, personalized whole-cell kinetic models of erythrocyte metabolism … based on fasting-state plasma and erythrocyte metabolomics and whole-genome genotyping,” the authors wrote. “We show that personalized kinetic rate constants, rather than metabolite levels, better represent the genotype. Additionally, changes in erythrocyte dynamics between individuals occur on timescales of circulation, suggesting detected differences play a role in physiology.”

While the authors cautioned that their small, proof-of-concept study will need to be confirmed by larger studies, they pointed out that their work demonstrated the feasibility of personalized kinetic models. The authors emphasized, for example, that their modelling efforts not only identified individuals at risk for a drug side effect (ribavirin-induced anemia), but also uncovered a potential protective mechanism (a genetic variation associated with inosine triphosphatase deficiency.

“We’re not just interested in predicting the efficacy of a drug, but its side effects as well,” said Bernhard Palsson, Ph.D., the Galetti Professor of Bioengineering at the Jacobs School of Engineering at UC San Diego.

“There needs to be a good way to obtain data about a drug’s side effects before exposing a lot of people to the drug. This predictive model could be used to figure out what these side effects are ahead of time,” added UC San Diego alumnus Aarash Bordbar, who did this research while a Ph.D. student in Dr. Palsson’s Systems Biology Research Group.

The researchers stressed that predictive models such as theirs would be extremely useful for pharmaceutical companies during the drug development stage. For example, pharmaceutical companies could conduct predictive screenings for drugs before clinical trials and determine which groups of patients would experience side effects and which ones wouldn’t.

“This study is a step forward in demonstrating that patients could be precisely treated based on their genetic makeup,” noted Dr. Palsson.

As a next step, researchers are also looking to develop predictive models for platelet cells, which are more complex than red blood cells. The ultimate goal is a liver cell model, researchers said, because the liver is where the majority of drugs are metabolized and where many drug side effects are manifested.