AI-driven project identifies up to 390 potential drugs against COVID
Joaquín Dopazo is leading a project that uses a machine learning algorithm to identify proteins involved in the illness and cross match this information with a database of drugs used to treat other conditions. “We checked if some of the proteins detected are therapeutic targets of other medications approved for other uses, and we concluded that these treatments could act on the mechanisms that COVID-19 uses to cause harm,” says the researcher, who in 2018 won a BBVA Foundation team grant in the Big Data category.
According to Joaquín Dopazo, Director Clinical Bioinformatics Area Fundación Progreso y Salud, in Seville , they are using a “mathematical model of the mechanism of the illness”, which provides a dynamic representation of what COVID-19 patients go through. “Our goal is to apply the same system we already apply to identify the therapeutic targets against rare diseases in the project that the BBVA Foundation backed,” he says.
The project analyzes the disease’s process as a whole, from the virus’ infection mechanism to the reactions it triggers in the organism. Something very relevant, especially because “people are dying not from the infection itself, but as a result of the brutal immunologic response caused by the virus, which triggers a lethal inflammatory process.”
Joaquín Dopazo (second from the left), with the research team - Fundación BBVA
Although still preliminary, the machine learning solution has allowed them to identify about 390 potential drugs, that may be able to act on the virus’ therapeutic targets and the infection process. Safety, the four that are yielding better results, and which are already in the trial stage, are choloriquine and hydroxichloriquine, which are already used against malaria, oseltamivir (remdesivir), an antiviral drug, and tocilizumab (Actemra) an immunosupressant for rheumatoid arthritis. “What’s interesting is that the program has precisely recommended chloroquine and hydroxychloroquine, which evidences the reliabillity of the process, and the other drugs that it is recommending may be valid,” says Dopazo.
The countdown to find a treatment
This research project attests to the tremendous biomedical potential of machine learning as scientific tool. Thanks to the progress in this field, smart machines can be trained already to classify datasets with human-like accuracy, if not better. “This type of artificial intelligence can help you detect patterns and relationships very efficiently. It’s like a GPS that guides you as you search and allows you to aim much more accurately when trying to find potential therapeutic targets,” says Dopazo. For the researcher, “the machine doesn’t do anything a human being wouldn’t be able to, but it is very useful because it helps detect cause-effect links without biases, which is what people typically do.”
This way, the artificial intelligence solution can help identify much faster whether any of the drugs in use to treat other diseases can be used against COVID-19. This would imply significant cost savings and, more importantly, shortening the time needed to come up with an effective treatment in just a few months.