Along with the science sometimes comes an inspirational call to arms. “We can bring this epidemic under control,” he wrote in a thread that was retweeted 5,000 times. “This is the Apollo program of our times. Let's get to it.”

In his 19-part March 18 Twitter thread, Bedford offers way to do just that. One path out of the crisis, he says, could be via a massive effort to roll out in-home testing kits and drive-through sites to spot cases early on and then combine those with cellphone location data to trace all the previous movements of those who test positive.

He says he finds his newfound Twitter fame a bit bewildering. “This has been very, very surreal,” says Bedford, who's been working 16-hour days since the outbreak started. “I am getting all this attention for doing this, and meanwhile everyone else's lives are being upended in terrible ways.”

One of his key collaborators, Richard Neher, is a computational biologist at the University of Basel. Neher says the two scientists hit upon the idea of tracking virus evolution in real time using an interactive website after meeting at a conference at the University of California Santa Barbara in 2014. Their original idea was focused on influenza evolution, with the goal of helping vaccine makers predict which strains are likely to spread around the world in the next flu season. But over time their website, Nextstrain.org, evolved to include data from multiple outbreaks including Zika, Enterovirus D68 and Ebola.

When the coronavirus hit, Bedford and Neher had customized software ready to roll for rapidly analyzing hundreds of virus genomes. “We hit the ground running here because all of this basic infrastructure was in place,” Neher says.

Since then, Nextstrain has become a 24/7 operation, staffed with researchers at Bedford’s and Neher’s labs in Seattle and Basel, along with another scientist in New Zealand. With global coverage, someone is always on call to start analyzing data as soon as a new viral genome is released to gisaid.org, a website where scientists are posting the information. It takes about 20 to 30 minutes to analyze a new viral genome, allowing the website to be updated frequently.

Bedford sees his work as expanding, not replacing, the utility of existing virus-tracing methods, providing new data streams to complement traditional epidemiology.  And while the evidence he gathers stops short of proving a chain of transmission, “my suspicion is almost everything we have seen in the Seattle area is part of the same transmission chain,” he says.

He started analyzing coronavirus genomes from China as soon as they began to flow into public databases on January 10th. At the time, health authorities were claiming that the virus had limited ability to spread between people. But Bedford found something alarming: The viral genomes were too similar to derive from viruses from different animals infecting people on multiple occasions. Instead, the genome data suggested that someone had acquired it from a single infected animal around early December — and it had been spreading from person to person ever since.

“This genomic data represented one of the first and strongest indications of sustained epidemic spread,” Bedford said in a Jan. 31 blog post. “I spent the week of Jan 20 alerting every public health official I know.”

Bedford and Neher are limited by the amount of genome data that is available. So far almost 1,000 patients have had their viral genomes analyzed, out of more than 350,000 people who have been infected. There are few virus genome sequences from New York, which has surpassed Washington as the hardest-hit state in the country.  Overwhelmed testing centers often don't have manpower to spare to do genome analysis when so many people are having trouble getting test results.