Fighting COVID-19 with the Synergistic Rhythms of the Tango
IBM Fellow and Argentinian Gustavo Stolovitzky knows a thing or two about the tango. And the famous Buenos Aires social dance may help us to address a vexing question for life sciences — what happens when two medications are applied to the same cell?
The answer may just be found in the synergy of twos.
Stolovitzky and his team have been investigating this question for the past five years at the Icahn School of Medicine at Mount Sinai in East Harlem New York, where Stolovitzky was an Adjunct Professor until last year. The problem is a challenging one — mainly because it doesn’t seem to agree with the basic laws of physics.
“If you have two forces applied to a body, the resulting force is the sum of the two forces, a principle known as superposition. This is well understood in physics, but it doesn’t work at the transcriptome level when applying two different drugs on the same cell,” says Stolovitzky.
This is why, say, HIV patients are treated with a drug cocktail that combines drugs that are individually ineffective in stopping the virus. In combination though, they can lead to dramatic improvements. Stolovitzky wants to understand drug combinations and thinks that it can be achieved using the right data, machine learning and supercomputing.
In his recently published paper in eLife “The transcriptomic response of cells to a drug combination is more than the sum of the responses to the monotherapies,” he and his team looked at drug combinations for breast and prostate cancer.
Now they are using the IBM Summit supercomputer at Oak Ridge National Lab to run the same algorithm for a different purpose: testing drug combinations in an effort to help fight coronavirus.
Many things in life come down to synergy — and in this research, synergy means how cells respond to combinations of drugs. In the paper, the scientists studied the synergy between Tamoxifen, a selective estrogen receptor modulator used to prevent breast cancer, and Mefloquine, an antimalarial medicine. The results, says Stolovitzky, were surprising.
While there were little changes to the gene expression in isolation, when they put the drugs together, the gene expression change was massive. It increased in proportion with the synergies that were observed at the cell culture.
Using this concept, the team found that drugs can be chosen to be synergistic — meaning that they are more potent in killing cancer cells when combined rather than what would be expected from the response to single drugs in isolation. This tend to occur when the gene expression of the two drugs is correlated at the molecular level, says Stolovitzky, even though they are entirely different in their mechanism of action.
“It is as if at the level of each gene they are stubborn and they don’t want to change by the effect of a single drug,” says Stolovitzky. “But when two drugs are synergistic, the genes are ‘pushed’ at the same time and in the same direction, forcing massive synergistic gene expression changes that activate death in cancer cells — and is critical for cancer patients.”
Based on this idea, the scientists created an algorithm that finds drugs that elicit correlated gene expression responses, as well as genes that have a better response to combination of certain drugs. Working with his former student Jennifer Diaz from the Icahn School of Medicine at Mount Sinai, Stolovitzky is now in the process of scaling the algorithm using IBM’s Summit supercomputer, available thanks to the COVID-19 High Performance Computing Consortium.
Stolovitzky and Diaz applied it to a database of FDA-approved drugs to find synergistic activation of biological processes important in the response of cells to the SARS-COV-2 virus that causes COVID-19. These processes include glycosylation, proteosome, endosome, autophagy, and protein trafficking — and they want to determine which synergistic combinations are best.
To do so, the first step will be creating a database of over 700,000 drug combinations. For each combination, the computer predicted synergistic gene expression using an algorithm trained on an independent drug synergy dataset from a different cell line. This algorithm was then applied to the three cell lines studied in the Connectivity Map, a public database.
Running this algorithm, says Stolovitzky, yielded more than 2.8 million sets of predictions. They then selected the most potentially promising combinations, which will soon be experimentally tested by Mount Sinai researchers to determine how effective they are in killing cell lines infected with the coronavirus — in a bid to help halt the ongoing pandemic.
It really does take two to tango, even at the molecular level.