Instead of trying to work through these issues at the national level, the sequencing contracts allow individual public health agencies to request the names and contact information of people who have tested positive for variants of concern. But that just pushes the same problems of data ownership down the chain.
“Some states are very good and want to know a lot about variants that are circulating in their state,” says Labcorp’s Brian Krueger. “The other states are not.”
Public health epidemiologists often have little experience with bioinformatics, using software to analyze large datasets like genomic sequences. Only a few agencies have pre-existing sequencing programs; even if they did, having each jurisdiction to analyze just a small slice of the dataset undercuts how much knowledge can be gleaned about real-world behavior.
Getting around those issues—making it easier to connect sequences and clinical metadata on a large scale—would require more than just root and branch reform of privacy regulations, however. It would need a reorganization of the entire healthcare and public health systems in the US, where each of the 64 public health agencies operate as fiefdoms, and there is no centralization of information or power.
“Metadata is the single biggest uncracked nut,” says Jonathan Quick, managing director of pandemic response, preparedness, and prevention at the Rockefeller Foundation. (The Rockefeller Foundation helps fund coverage at MIT Technology Review,, although it has no editorial oversight.) Because it’s so hard for public health to put together big enough datasets to really understand real-world variant behavior, our understanding has to come from vaccine manufacturers and hospitals adding sequencing to their own clinical trials, he says.
It’s frustrating to him that so many huge datasets of useful information already exist in electronic medical records, immunization registries, and other sources, but can’t easily be used.
“There’s a whole lot more that could be learned, and learned faster, without the shackles we put on the use of that data,” says Quick. “We can’t just rely on the vaccine companies to do surveillance.”
Boosting state-level bioinformatics
If public health labs are expected to focus more on tracking and understanding variants on their own, they’ll need all the help they can get. Doing something about variants case-by-case, after all, is a public health job, while doing something about variants on a policy level is a political one.
Public health labs generally use genomics to expose otherwise-hidden information about outbreaks, or as part of track and trace efforts. In the past, sequencing has been used to connect E. coli outbreaks to specific farms, identify and interrupt chains of HIV transmission, isolate US Ebola cases, and follow annual flu patterns.
Even those with well-established programs tend to use genomics sparingly. The cost of sequencing has dropped precipitously over the last decade, but the process is still not cheap, particularly for cash-strapped state and local health departments. The machines themselves cost hundreds of thousands of dollars to buy, and more to run: Illumina, one of the biggest makers of sequencing equipment, says labs spend an average of $1.2 million annually on supplies for each of its machines.
“We’ll miss a ton of opportunities if we just give health departments money to set up programs without having a federal strategy so that everyone knows what they’re doing”
Health agencies don’t just need money; they also need expertise. Surveillance requires highly trained bioinformaticians to turn a sequence’s long strings of letters into useful information, as well as people to explain the results to officials, and convince them to turn any lessons learned into policy.
Fortunately, the OAMD has been working to support state and local health departments as they try to understand their sequencing data, employing regional bioinformaticians to consult with public health officers and facilitating agencies’ efforts to share their experiences.
It is also pouring hundreds of millions into building and supporting those agencies’ own sequencing programs—not just for covid, but for all pathogens.
But many of those agencies are facing pressure to sequence as many covid genomes as possible. Without a cohesive strategy for collecting and analyzing data, it’s unclear how much utility those programs will have.
“We’ll miss a ton of opportunities if we just give health departments money to set up programs without having a federal strategy so that everyone knows what they’re doing,” says Warmbrod.
Initial visions, usurped
Mark Pandori is director of the Nevada state public health laboratory, one of the programs OAMD supports. He has been a strong proponent of genomic surveillance for years. Before moving to Reno, he ran the public health lab in Alameda County, California, where he helped pioneer a program using sequencing to track how infections were being passed around hospitals.
Turning sequences into usable data is the biggest challenge for public health genomics programs, he says.
“The CDC can say, ‘go buy a bunch of sequencing equipment, do a whole bunch of sequencing.’ But it doesn’t do anything unless the consumers of that data know how to use it, and know how to apply it,” he says. “I’m talking to you about the robotics we need to get things sequenced every day, but health departments just need a simple way to know if cases are related.”
When it comes to variants, public health labs are under many of the same pressures the CDC faces: everyone wants to know what variants are circulating, whether or not they can do anything with the information.
Pandori launched his covid sequencing program hoping to cut down on the labor needed to investigate potential covid outbreaks, quickly identifying whether cases caught near each other were related or coincidental.
His lab was the first in North America to identify a patient reinfected with covid-19, and later found the B.1.351 variant in a hospitalized man who had just come back from South Africa. With rapid contact tracing, the health department was able to prevent it from spreading.
A new training model, dubbed “KnowNo,” aims to address this problem by teaching robots to ask for our help when orders are unclear. At the same time, it ensures they seek clarification only when necessary, minimizing needless back-and-forth. The result is a smart assistant that tries to make sure it understands what you want without bothering you too much.
Andy Zeng, a research scientist at Google DeepMind who helped develop the new technique, says that while robots can be powerful in many specific scenarios, they are often bad at generalized tasks that require common sense.
For example, when asked to bring you a Coke, the robot needs to first understand that it needs to go into the kitchen, look for the refrigerator, and open the fridge door. Conventionally, these smaller substeps had to be manually programmed, because otherwise the robot would not know that people usually keep their drinks in the kitchen.
That’s something large language models (LLMs) could help to fix, because they have a lot of common-sense knowledge baked in, says Zeng.
Now when the robot is asked to bring a Coke, an LLM, which has a generalized understanding of the world, can generate a step-by-step guide for the robot to follow.
The problem with LLMs, though, is that there’s no way to guarantee that their instructions are possible for the robot to execute. Maybe the person doesn’t have a refrigerator in the kitchen, or the fridge door handle is broken. In these situations, robots need to ask humans for help.
KnowNo makes that possible by combining large language models with statistical tools that quantify confidence levels.
When given an ambiguous instruction like “Put the bowl in the microwave,” KnowNo first generates multiple possible next actions using the language model. Then it creates a confidence score predicting the likelihood that each potential choice is the best one.
The news: A new robot training model, dubbed “KnowNo,” aims to teach robots to ask for our help when orders are unclear. At the same time, it ensures they seek clarification only when necessary, minimizing needless back-and-forth. The result is a smart assistant that tries to make sure it understands what you want without bothering you too much.
Why it matters: While robots can be powerful in many specific scenarios, they are often bad at generalized tasks that require common sense. That’s something large language models could help to fix, because they have a lot of common-sense knowledge baked in. Read the full story.
—June Kim
Medical microrobots that travel inside the body are (still) on their way
The human body is a labyrinth of vessels and tubing, full of barriers that are difficult to break through. That poses a serious hurdle for doctors. Illness is often caused by problems that are hard to visualize and difficult to access. But imagine if we could deploy armies of tiny robots into the body to do the job for us. They could break up hard-to-reach clots, deliver drugs to even the most inaccessible tumors, and even help guide embryos toward implantation.
We’ve been hearing about the use of tiny robots in medicine for years, maybe even decades. And they’re still not here. But experts are adamant that medical microbots are finally coming, and that they could be a game changer for a number of serious diseases. Read the full story.
We haven’t always been right (RIP, Baxter), but we’ve often been early to spot important areas of progress (we put natural-language processing on our very first list in 2001; today this technology underpins large language models and generative AI tools like ChatGPT).
Every year, our reporters and editors nominate technologies that they think deserve a spot, and we spend weeks debating which ones should make the cut. Here are some of the technologies we didn’t pick this time—and why we’ve left them off, for now.
New drugs for Alzheimer’s disease
Alzmeiher’s patients have long lacked treatment options. Several new drugs have now been proved to slow cognitive decline, albeit modestly, by clearing out harmful plaques in the brain. In July, the FDA approved Leqembi by Eisai and Biogen, and Eli Lilly’s donanemab could soon be next. But the drugs come with serious side effects, including brain swelling and bleeding, which can be fatal in some cases. Plus, they’re hard to administer—patients receive doses via an IV and must receive regular MRIs to check for brain swelling. These drawbacks gave us pause.
Sustainable aviation fuel
Alternative jet fuels made from cooking oil, leftover animal fats, or agricultural waste could reduce emissions from flying. They have been in development for years, and scientists are making steady progress, with several recent demonstration flights. But production and use will need to ramp up significantly for these fuels to make a meaningful climate impact. While they do look promising, there wasn’t a key moment or “breakthrough” that merited a spot for sustainable aviation fuels on this year’s list.
Solar geoengineering
One way to counteract global warming could be to release particles into the stratosphere that reflect the sun’s energy and cool the planet. That idea is highly controversial within the scientific community, but a few researchers and companies have begun exploring whether it’s possible by launching a series of small-scale high-flying tests. One such launch prompted Mexico to ban solar geoengineering experiments earlier this year. It’s not really clear where geoengineering will go from here or whether these early efforts will stall out. Amid that uncertainty, we decided to hold off for now.
