Machine learning algorithms can help physicians with diagnoses and can detect precancerous lesions or cancer in the early stages, which could easily be missed, resulting in faster diagnosis. Early diagnosis and treatment are essential for preventing morbidity and mortality. Determining the treatment regimen that is most likely to be effective is also important. Not all patients will respond to well to certain treatments and predicting likely outcomes remains a challenge, so too determining which combinations of cancer drugs should be used.
A team of researchers from the University of Helsinki and the University of Turku in Finland have developed a machine learning algorithm capable of predicting how different combinations of cancer drugs will kill cancerous cells
The researchers used data from previous studies investigating the association between drugs and cancer cells to train their machine learning framework – comboFM, including studies on patient-derived cells and cell lines.
“comboFM models the cell context-specific drug interactions through higher-order tensors, and efficiently learns latent factors of the tensor using powerful factorization machines. The approach enables comboFM to leverage information from previous experiments performed on similar drugs and cells when predicting responses of new combinations in so far untested cells; thereby, it achieves highly accurate predictions despite sparsely populated data tensors,” explained the researchers.
Using data from cancer cell line pharmacogenomic screens, the researchers showed their framework had a high predictive performance in various scenarios. Using anticancer drug combination response data from the Nation Cancer Institute-ALMANAC study, the researchers combined 50 unique FDA-approved drugs in 617 combinations, which were screened in different concentration pairs across 60 cell lines originating from nine tissue types.
The Nation Cancer Institute-ALMANAC data includes a total of 333,180 drug combination response measurements and 222,120 monotherapy response measurements of single drugs against percentage growth of the cell lines.
“The model gives very accurate results. For example, the values of the so-called correlation coefficient were more than 0.9 in our experiments, which points to excellent reliability,” said Juho Rousu, PhD, professor, computer science, Aalto University. Juho Rousu, PhD, professor, computer science, Aalto University.
The machine learning algorithm was shown to be reliable and can be used by cancer researchers to determine the best possible drug combinations, ratios, and concentrations for further research.
“Given the high cost of the experimental screening of drug combinations, comboFM has the potential to provide time- and cost-effective means toward prioritizing the most promising drug combinations for further preclinical or clinical studies,” said the researchers.
You can read more about the study in the paper – Leveraging multi-way interactions for systematic prediction of pre-clinical drug combination effects – which was recently published in the Nature Communications. DOI: 10.1038/s41467-020-19950-z