Computer Model Uses Machine Learning to Support Cancer Therapy

Researchers have recently created a machine learning computer model that can simulate the metabolism of cancer cells. This team, from the Life Sciences Research Unit at the University of Luxembourg, used this technique to analyze the effects of various drugs on stopping cancer development. Their findings were covered in an open access article in the Lancet journal EBioMedicine.

Once cells undergo morphological changes to become malignant, they begin utilizing different metabolic processes than healthy cells. Though this metabolic pathway is focused acutely on growth and proliferation, this limitation makes the cancer cell less able to adapt to changes.

“Their metabolism is much leaner than that of healthy cells, as they are just focused on growth,” explained Thomas Sauter, Professor of Systems Biology at the University of Luxembourg and lead author of the paper. “However, this makes them more vulnerable to interruptions in the chain of chemical reactions that the cells depend on. Whereas healthy cells can take alternative routes when one metabolic path is disabled, this is more difficult for cancer cells. In our study, we investigated how drugs or combinations of drugs could be used to switch off certain proteins in cancer cells and thereby interrupt the cell’s metabolism.”

After creating digital models of both healthy and cancerous cells, the researchers then incorporated genetic data from 10,000 patients within the Cancer Genome Atlas of the American National Cancer Institute. These computer models were used by the researchers to simulate the effects of different compounds on cellular metabolism. Simulating with the cancer cell models showed which drugs can effectively inhibit cancer growth, while simulating with the healthy cells tested the safety of these drugs. By screening for drugs that are effective and safe, this system allows researchers to rule out poor candidates before lab testing.

The researchers assessed roughly 800 drugs on these cell models using their predictive machine learning model and 40 of these tested compounds were found to combat cancer development. About half of these drugs were already identified as anti-cancer therapies, but only 17 are currently approved only for other treatments.

“Our tool can help with the so-called ‘drug repositioning’, which means that new therapeutically purposes are found for existing medication,” Sauter said. “This could significantly reduce the cost and time for drug development.”

The models tested so far have only been for colorectal cancer, but Sauter claims that the AI system would work for all types of cancer as well. He and his colleagues will potentially move into developing commercial uses for this technique.

First author Dr. Maria Pacheco highlighted the efficiency of their mathematical approach. “We managed to create 10,000 patient models within one week, without the use of high-performance computing. This is exceptionally fast,” explained Pacheco, who is also a postdoctoral researcher at the University of Luxembourg.

Another collaborator on the study, Dr. Elisabeth Letellier, feels that their work could revolutionize how individualized cancer therapy regimens are designed. “In the future, this could allow us to build models of individual cancer patients and virtually test drugs in order to find the most efficient combination. This could also bring fresh hope to patients for whom known therapies haven proven to be ineffective,” said Letellier, principal investigator at the Molecular Disease Mechanisms group at the University of Luxembourg.