Jake Lever spotted a gap in the market midway through two years of postdoctoral research in the bioengineering department at Stanford University.
Noticing last April that his colleagues were struggling to sift through the mountains of literature published on the coronavirus every day, he put his skills as a “biomedical text miner” to work. Four months later, CoronaCentral.ai was born.
The website, which has since received funding from Chan Zuckerberg BioHub research center, uses deep learning, a subset of artificial intelligence, to compile and categorize published Covid-19 articles and preprints.
” There is so much [information] there it was a huge challenge to find what you want, ”Lever said. But by putting it all in different topics and tying it to esteem metrics, it’s easier to see what’s being read, he added.
CoronaCentral can now rely on artificial intelligence, but it took a lot of human work to begin with. To “train” the system, Lever spent weeks sorting approximately 3,200 research articles into categories such as “journals,” “epidemiology” and “meta-analysis.”
“Language processing systems work best when you provide them with concrete and benchmark examples of what you want,” he said. “You have a lot of coffee, you sit down and you just start. “
Trends are emerging among the more than 150,000 articles gathered by the website since the start of the pandemic. Predictive research using mathematical models to predict the spread of the virus in different settings has declined as the focus is placed on analyzing the effects of long Covid.
Articles on how to adapt medical practice during the pandemic are also common, Lever said. “So many people write things like ‘we’re doing colorectal cancer surgery, how should we do it now?'”
As scientists are “increasingly required to undertake interdisciplinary research,” Lever believes websites such as CoronaCentral will play an important role in how they digest the latest findings.
“With the pace of research, it is impossible for a single scientist to keep up to date in many areas,” he said.
Last month, Lever traded California for Glasgow for a teaching position at the college’s computer science school, near where he grew up. His doctorate was in bioinformatics, a topic he describes as “using computers to analyze DNA sequences, spot patterns, identify mutations or, in the case of Covid, analyze variants.”
Sequence analysis has become “so ubiquitous and easy and convenient to do that we are able to study variants as we could not have done 10 years ago,” Lever added. “In my eyes, we are in a very lucky place. “
This is the 10th article in a series for the blog that explores the effects of the pandemic on people and businesses around the world.