COVID-19 brought unforeseen challenges to the pharmaceutical industry. The pressure was on to condense years worth of clinical research into mere weeks and months. A vaccine for the novel coronavirus was of utmost precedence, and scientists had no choice but to turn to automated solutions for help, including various forms of artificial intelligence. Data scientists at Pfizer must have breathed sighs of relief when they realized that the clinical trial data that would have taken them over a month to process was prepared in just twenty-two hours hours with the help of the Smart Data Query (SDQ) machine learning tool. To put that into perspective, that’s an over 96% reduction in processing time. It’s AI tools like this that helped put vaccines out for testing in March of 2020, just a few months after the initial COVID-19 outbreak. Such speed is an enormous feat in immunization history when considering that the shortest vaccine development time record was originally set by the mumps vaccine, which began market distribution in about four years.
There’s no denying that AI’s role in vaccination development has advanced greatly in the past several years. In addition to clinical trial data processing, AI tools have been used by researchers to ascertain the structure of the SARS-CoV-2 virus and help them predict the components that will elicit immune responses. Moreover, genetic mutations can be tracked with such AI tools, ultimately helping scientists determine the efficacy of a vaccine in the future. At Moderna, for instance, another major COVID-19 vaccine supplier in recent years, scientists increased their small-scale monthly mRNA output thirty-three fold with the help of robotic automation, digital systems, and process automation that worked with AI algorithms. Their algorithms are able to help find optimal paths for coding protein in their mRNA sequence design. As Dave Johnson, chief data and AI officer at Moderna, points out, their AI algorithms have been integrated into their systems to the point where scientists can now “just press a button and the work is done for them.”
Hearing Dave Johnson utter those words is both an exciting but glooming prospect. While the automation is relieving massive amounts of burden on data scientists, such human withdrawal from the process could possibly open room for errors and biases to crop in the final product if the algorithms aren’t properly regulated. When asked how the scientists at Moderna feel that some of their vaccine development work is being replaced by AI, Johnson confirms that although his organization “believes in giving people a lot of responsibility,” they are seeing a positive impact on the morale, happiness, and collaboration capabilities of their teams as a result of having AI help with the more repetitive processes that they do. Johnson went on to further stress how limited the set of responsibilities that their AI has by reassuring people that no AI in the pharmaceutical industry can eradicate the entire drug discovery process. He claimed that such a full-scale job, which would involve predicting the efficacy of a certain drug from the structure of just a molecule, is “completely unrealistic.”
All in all, Johnson sums up quite well how AI has been used at Moderna in the course of a roughly twenty-minute interview with folks at MITSloan Management review. For more details about the use of AI in vaccine production, check out the transcript of the interview at https://sloanreview.mit.edu/audio/ai-and-the-covid-19-vaccine-modernas-dave-johnson/. Though, vaccine development isn’t the last place where “AI is used for AI.” In April of 2021, when hopes of curbing the pandemic through mass vaccination were high, AI-powered algorithms were in use by large healthcare organizations for handling, storing, and tracking the outcome of vaccines. Although perhaps not as widely used at the time, AI utilities for vaccine distribution had the capabilities of ensuring mass vaccination through automatic strategic supply chain adjustments, as David Smith, associate Vice President of virtual medicine at UMass Memorial Health Care, pointed out.
It’s very exciting to think that AI is being used to develop, test, and distribute vaccines. Some well-placed skepticism is not out of the norm though given prior ethical dilemmas involving AI in healthcare, such as the one discussed in the second blog post of this series. I’ll say this time and time again: good regulation and adherence to new standards will provide for a more ethical AI future. At the very least you can rest assured knowing that the AI used in the COVID-19 vaccine process will not “alter [your] DNA and hook [you] to an AI” (I’m not even going to explain this one, I’m just going to put a link here: https://towardsdatascience.com/will-the-covid-19-vaccine-alter-my-dna-and-hook-me-to-an-ai-f7d086b94e07)
- Jon Cili
Comentarios