Dec. 1 (UPI) — Researchers in Finland have developed a machine-learning tool that accurately predicts how different cancer drugs kill disease cells, according to a paper published Tuesday by Nature Communications.
The new artificial intelligence-based platform uses data from published studies to calculate the potential effectiveness of various treatments or combinations of treatments in individual patients, the researchers said.
It could prove useful in helping physicians ideally tailor cancer treatment for individual tumors, they said.
In addition, “This [AI tool] will help cancer researchers to prioritize which drug combinations to choose from thousands of options for further research,” researcher Tero Aittokallio said in a press release.
Aittokallio, a professor at the University of Helsinki’s Institute for Molecular Medicine Finland, helped develop the platform.
When healthcare professionals treat patients who suffer from advanced cancers, they usually need to use a combination of different therapies, including chemotherapy drugs, radiation therapy or both, according to the Finnish researchers.
Combining chemotherapy drugs often improves the effectiveness of treatment while reducing harmful side effects.
However, experimental screening of drug combinations can be very slow and expensive, they said.
For this study, the researchers tested their AI-based model using data on more than 330,000 potential treatment combinations, including U.S. Food and Drug Administration-approved chemotherapy drugs.
The model found associations between drugs and cancer cells that were not observed previously.
The finding, researchers said, is based on a correlation coefficient, a statistical method measuring how strong a relationship is between two variables — in this case, drug combinations and cancer cell death.
Based on this analysis, the model “gives very accurate results … which points to excellent reliability,” said co-author Juho Rousu, a professor at Aalto University.
This indicates that the new machine learning method can help quickly identify best combinations to selectively kill cancer cells with specific genetic or functional makeup, potentially increasing effectiveness, they said.