Researchers at Vanderbilt University Medical Center have developed a new method for identifying drugs for the repurposing trials that can lead to new indications for drugs already in use.
The method combines gene expression patterns of diseases and drugs with patient data from electronic health records (EHRs). The team’s proof-of-concept report appeared Jan. 10 in Nature Communications.
“Compared to the drug discovery and development route, repurposing can be a much quicker, more economical way to bring new therapies to bear on patient problems,” said Wei-Qi Wei, MD, PhD, assistant professor of Biomedical Informatics, who led the project with Patrick Wu, an MD/PhD student who works in Wei’s lab.
“With the safety of drugs already at market tending to be well established,” said Wu, “what remains is to establish a given drug candidate’s efficacy for the new indication. Our method aims to help teams identify the absolute best candidates for those efficacy trials.”
Diseases come with gene expression signatures, providing a window into their molecular workings. Drugs, meanwhile, are seen as having gene expression signatures of their own.
According to the report, an emerging principle in the repurposing field holds that when a disease and a drug happen to push gene expression in opposite directions, it’s a hint that the drug might be effective in treating the disease. This supplies the starting point for the method set out by Wei, Wu and colleagues.
The team focused on some 21,000 drug compounds, and two all too common diseases, hyperlipidemia (high cholesterol) and hypertension. To draw disease-drug gene expression comparisons, the team used published genomic results and various public databases — MetaXcan, S-PrediXcan, iLINCS — arriving at 149 candidate drugs for hyperlipidemia and 178 candidates for hypertension.
The team next used VUMC’s database of 3.2 million deidentified EHRs, the so-called synthetic derivative, to check whether, based on drug exposure documentation and test results for cholesterol and blood pressure, any of the candidate drugs identified based on gene expression appeared to confer protection against hyperlipidemia or hypertension. A bit more than half of the repurposing candidates captured in the gene expression scan turned out to meet inclusion criteria for the EHR scan.
Among 178 drugs viewed against test results in the EHR, 12 emerged as lowering blood cholesterol, 23 as lowering blood pressure. Among these 35 drugs were 10 already used to treat the two diseases in question. “These data signals that were returned for established cholesterol- and blood pressure-lowering drugs, first in terms of gene expression and then in terms of clinical test results, would appear to support our approach of combining the two distinct data types,” Wu said.
To confirm their findings, the team next turned to data from a National Institutes of Health (NIH) initiative, the All of Us Research Program, established in 2016 (the program’s Data and Research Center, led by Paul Harris, PhD, happens to be housed at VUMC). The initiative’s growing database, called the Researcher Workbench, provides EHR data on more than 236,000 participants from around the country.
The team used All of Us to test the signals returned from the synthetic derivative, 12 for hyperlipidemia, 23 for hypertension. In All of Us, one repurposing candidate for hyperlipidemia was shown to lower cholesterol (along with four drugs already established for the indication), and four repurposing candidates for hypertension were shown to lower blood pressure (along with two established drugs for this indication).
Treatment effects found in All of Us echoed those that had been seen in the synthetic derivative.
“For the repurposing candidates returned in this proof of concept, the effects that we measured on cholesterol and blood pressure happen to be small and not clinically significant,” Wei said. “Meanwhile, there are other diseases that lack effective drug treatments altogether. We’ve set out a significantly more sensitive and rigorous method for finding drugs for repurposing trials, and we look forward to expanding upon this approach and to other teams using our approach and statistical tool kit in the search for new drug repurposing candidates.”
Also on the study from VUMC were QiPing Feng, PhD, Vern Kerchberger, MD, Scott Nelson, PharmD, MS, Qingxia Chen, PhD, Bingshan Li, PhD, Todd Edwards, PhD, MS, Nancy Cox, PhD, Elizabeth Phillips, MD, C. Michael Stein, MBChB, Dan Roden, MDCM, and Joshua Denny, MD, MS.
The study was supported by the NIH (HL133786, TR000445).
