Previous articles posted on this blog have focused on different methods that Artificial Intelligence has used to combat the ongoing coronavirus pandemic: from detection to prediction models. As it turns out, AI can be used to help select the right drug(s) to repurpose them in the fight against COVID-19.
Sorting out the best available solutions
There is no available vaccine for the coronavirus, despite an ongoing, worldwide research effort to develop one. Therefore, the medical field has turned into testing out all the available drugs to discover common traits that have in the past had success against similar epidemic outbreaks. Deemed a faster route to solving the crisis than waiting for the launch of a new drug, scientists are making intelligent assumptions to pick a few candidates, basing their choice on research and plain intuition.
To this end, deep neural networks could help doctors find antivirals against a new target. Especially interesting is the way that the developed algorithms do not simply look for new drugs — it also scans through a database of chemicals, already licensed for certain diseases, which could also theoretically work against symptoms of coronavirus. Using virtual screening, AI scans the database of existing compounds that were utilized on related illnesses, and then singles out those with common traits. From there, the clinics that are responsible for treating those who are ill, decide which particular selected drugs should be tested and observed.
A preprint paper released by Konstantin Avchaciov, Olga Burmistrova, and Peter Fediche of Gero LTD described how deep neural networks could help doctors identify antivirals against the new pandemic. The team looked at similarities between the genes of COVID-19 and a similar virus that caused the SARS epidemic in the early 2000s, and through machine learning, were able to single out cells that were resilient to certain treatments with SARS, and those that were successful.
Due to the genetic similarity of COVID-19 and SARS viruses, the team were able to select those drugs that were successful, in order to test them against the coronavirus.
We’d like to add that here at Alphamoon we have worked on virtual screening methods using graph-based convolutional neural networks, the link to the publication of the research can be found here.
Why this matters
The usage of AI for drug repurposing is not new, as countless studies and tests have been performed on the topic in recent years.
When compared to the capabilities of human researchers, AI has the potential to look further into the results of medications at the chemical or genetic stage. Whereas a person would take a long time to look through the smallest details when comparing drugs on the microscopic level, AI virtual screening will not only do a quick scan, but will detect the necessary information with precision.
Due to the accuracy of virtual screening on the basis of previous viruses and diseases used against them, medical personnel can quickly find drugs that could be carefully tested to treat those infected. Drug repurposing is sometimes seen as a last-ditch effort, but it can yield a very potent way of halting any ongoing pandemic.