BioPharma

Researchers develop computer modeling to predict how drugs work in the body

Researchers at Columbia University have developed an algorithm that can determine how an individual will react to a drug – paving the way for drugs with fewer side effects, or finding new therapeutic uses for existing medications.

Columbia University researchers have developed an algorithm that shows the effects that drugs have on the body — potentially offering ways to develop more efficient drugs that produce fewer side effects. The results, published in Cell, could also pave the path for finding new therapeutic uses for existing drugs.

“For the first time we can perform a genome-wide search to identify the entire set of proteins that play a role in a drug’s activity,” study co-author Dr. Andrea Califano said in a statement.

It’s difficult to predict how a drug will interact with the body’s highly complex molecular setup, which is why side effects pop up. That’s why our clinical trial process is so riddled with trial and error — as the study points out, scientists currently have no way to “identify the full repertoire of proteins that are affected by a drug.”

Sponsored Post

Physician Targeting Using Real-time Data: How PurpleLab’s Alerts Can Help

By leveraging real-time data that offers unprecedented insights into physician behavior and patient outcomes, companies can gain a competitive advantage with prescribers. PurpleLab®, a healthcare analytics platform with one of the largest medical and pharmaceutical claims databases in the United States, recently announced the launch of Alerts which translates complex information into actionable insights, empowering companies to identify the right physicians to target, determine the most effective marketing strategies and ultimately improve patient care.

The researchers have created a new platform called DeMAND — Detecting Mechanism of Action by Network Dysregulation — that uses computer modeling of the network of protein interactions in a diseased cell. It will combine modeling data with information from experiments to figure out how proteins are impacted by a drug.

It’s tested the accuracy of the modeled predictions of drug reactions with diffuse B-cell lymphoma cells and found that there’s been 70 percent accuracy. It also used modeling to find that unrelated drugs sulfasalazine and altretamine actually share some similarities in their biological impact. Altretamine, which is used for ovarian cancer, can actually be used in a similar manner as sulfasalazine to treat inflammatory disease, the algorithms predict.

These in silico approaches to drug development are gaining steam in the scientific community. For instance, University of Toronto researcher Brendan Frey just launched a startup, Deep Genomics, that uses deep learning and artificial intelligence to predict cellular processes — and ultimately map out the way disease manifests itself. Massachusetts upstart Berg is ahead of the curve in using computer modeling to bring tailored therapy to the clinic more expediently.

These companies could save the pharmaceutical industry billions in trial and error, should the algorithms prove fruitful and work consistently.

“DeMAND could accelerate the drug discovery process and reduce the cost of drug development by unraveling how new compounds work in the body,” Columbia researcher and study coauthor Mukesh Bansal said in a statement. “Our findings on altretamine also show that it can determine novel therapeutic applications for existing FDA-approved drugs.”