UCSD researchers designed a virtual cell model that forecasts how patients will react to drugs.

Scientists at the UCSD Moores Cancer Center successfully designed a virtual cell model that predicts brain cancer cell responses to various drug treatments. The model’s algorithm uses the genetic and molecular information from patient tumor cells to determine which drug would be most effective in combatting the cancer.

Researchers published results in the May 21 online edition of the Journal of Translational Medicine. Project scientists spearheaded the study under the direction of Director of Neuro-Oncology at Moores Cancer Center Dr. Santosh Kesari. The successful findings will further personalize cancer treatment, according to lead author Dr. Sandeep Pingle.

Currently, personalized cancer therapy suggests the use of “mouse avatars.” In this method, scientists implant patient tumor cells into live mouse subjects. Researchers then test various treatments on the mice to see which drug has the most positive effect. Pingle’s method, instead, virtually models human tumor cells and predicts an ideal treatment, without relying on variable results from live subjects.

“For every tumor, in order for treatment to be effective, treatment has to be tailored to the specific molecular nature of the tumor,” Pingle said. “We can get that information from genomics and proteomics.”

To customize treatment, researchers remove a cell sample from the tumor of a patient and extract specific metabolic information. The team’s algorithm takes this data and generates a virtual profile of the patient’s natural, healthy cells. Researchers can then simulate metabolic mutations of this healthy model to produce virtual cancerous cells. Finally, the model simulates various drug treatments upon these cells and predicts the most combative drug for the patient’s specific cancer type.

The study published on May 21 observed glioblastoma, a particularly aggressive form of brain cancer with just a 50-percent survival rate within the first 15 months of diagnosis. Treatment is difficult, as each glioblastoma tumor has a unique cell composition.

Despite this, when researchers compared the reactions of virtual cells to the actual reactions of their living culture cell counterparts, they found a 75-percent response accuracy. The study solely tested glioblastoma cells, but its algorithm can be used on any cancer.

“Our ultimate goal is to take this technology to the clinic to identify the best drugs for treating each individual cancer patient,” Kesari said in a May 21 UCSD Health Sciences press release.

Ideally, the team wants to test its model in clinical trials within the year. Meanwhile, researchers are pursuing the next step for the model. Currently, the algorithm only accurately predicts which single drug would most effectively combat tumor cells.

“Single drug therapy is not very effective; eventually, the tumor develops a resistance against these single drugs,” Pingle said. “We want to test combinations.”