The burgeoning field of brain mapping

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here. 

The human brain is an engineering marvel: 86 billion neurons form some 100 trillion connections to create a network so complex that it is, ironically, mind boggling.

This week scientists published the highest-resolution map yet of one small piece of the brain, a tissue sample one cubic millimeter in size. The resulting data set comprised 1,400 terabytes. (If they were to reconstruct the entire human brain, the data set would be a full zettabyte. That’s a billion terabytes. That’s roughly a year’s worth of all the digital content in the world.)

This map is just one of many that have been in the news in recent years. (I wrote about another brain map last year.) So this week I thought we could walk through some of the ways researchers make these maps and how they hope to use them.  

Scientists have been trying to map the brain for as long as they’ve been studying it. One of the most well-known brain maps came from German anatomist Korbinian Brodmann. In the early 1900s, he took sections of the brain that had been stained to highlight their structure and drew maps by hand, with 52 different areas divided according to how the neurons were organized. “He conjectured that they must do different things because the structure of their staining patterns are different,” says Michael Hawrylycz, a computational neuroscientist at the Allen Institute for Brain Science. Updated versions of his maps are still used today.

“With modern technology, we’ve been able to bring a lot more power to the construction,” he says. And over the past couple of decades we’ve seen an explosion of large, richly funded mapping efforts.

BigBrain, which was released in 2013, is a 3D rendering of the brain of a single donor, a 65-year-old woman. To create the atlas, researchers sliced the brain into more than 7,000 sections, took detailed images of each one, and stitched the sections into a three-dimensional reconstruction.

In the Human Connectome Project, researchers scanned 1,200 volunteers in MRI machines to map structural and functional connections in the brain. “They were able to map out what regions were activated in the brain at different times under different activities,” Hawrylycz says.

This kind of noninvasive imaging can provide valuable data, but “Its resolution is extremely coarse,” he adds. “Voxels [think: a 3D pixel] are of the size of a millimeter to three millimeters.”

And there are other projects too. The Synchrotron for Neuroscience—an Asia Pacific Strategic Enterprise,  a.k.a. “SYNAPSE,” aims to map the connections of an entire human brain at a very fine-grain resolution using synchrotron x-ray microscopy. The EBRAINS human brain atlas contains information on anatomy, connectivity, and function.

The work I wrote about last year is part of the $3 billion federally funded Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative, which launched in 2013. In this project, led by the Allen Institute for Brain Science, which has developed a number of brain atlases, researchers are working to develop a parts list detailing the vast array of cells in the human brain by sequencing single cells to look at gene expression. So far they’ve identified more than 3,000 types of brain cells, and they expect to find many more as they map more of the brain.

The draft map was based on brain tissue from just two donors. In the coming years, the team will add samples from hundreds more.

Mapping the cell types present in the brain seems like a straightforward task, but it’s not. The first stumbling block is deciding how to define a cell type. Seth Ament, a neuroscientist at the University of Maryland, likes to give his neuroscience graduate students a rundown of all the different ways brain cells can be defined: by their morphology, or by the way the cells fire, or by their activity during certain behaviors. But gene expression may be the Rosetta stone brain researchers have been looking for, he says: “If you look at cells from the perspective of just what genes are turned on in them, it corresponds almost one to one to all of those other kinds of properties of cells.” That’s the most remarkable discovery from all the cell atlases, he adds.

I have always assumed the point of all these atlases is to gain a better understanding of the brain. But Jeff Lichtman, a neuroscientist at Harvard University, doesn’t think “understanding” is the right word. He likens trying to understand the human brain to trying to understand New York City. It’s impossible. “There’s millions of things going on simultaneously, and everything is working, interacting, in different ways,” he says. “It’s too complicated.”

But as this latest paper shows, it is possible to describe the human brain in excruciating detail. “Having a satisfactory description means simply that if I look at a brain, I’m no longer surprised,” Lichtman says. That day is a long way off, though. The data Lichtman and his colleagues published this week was full of surprises—and many more are waiting to be uncovered.

Now read the rest of The Checkup

Another thing

The revolutionary AI tool AlphaFold, which predicts proteins’ structures on the basis of their genetic sequence, just got an upgrade, James O’Donnell reports. Now the tool can predict interactions between molecules. 

Read more from Tech Review’s archive

In 2013, Courtney Humphries reported on the development of BigBrain, a human brain atlas based on MRI images of more than 7,000 brain slices. 

And in 2017, we flagged the Human Cell Atlas project, which aims to categorize all the cells of the human body, as a breakthrough technology. That project is still underway. 

All these big, costly efforts to map the brain haven’t exactly led to a breakthrough in our understanding of its function, writes Emily Mullin in this story from 2021.  

From around the web

The Apple Watch’s atrial fibrillation (AFib) feature received FDA approval to track heart arrhythmias in clinical trials, making it the first digital health product to be qualified under the agency’s Medical Device Development Tools program. (Stat)

A CRISPR gene therapy improved vision in several people with an inherited form of blindness, according to an interim analysis of a small clinical trial to test the therapy. (CNN)

Long read: The covid vaccine, like all vaccines, can cause side effects. But many people who say they have been harmed by the vaccine feel that their injuries are being ignored.  (NYT)

Cancer vaccines are having a renaissance

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here. 

Last week, Moderna and Merck launched a large clinical trial in the UK of a promising new cancer therapy: a personalized vaccine that targets a specific set of mutations found in each individual’s tumor. This study is enrolling patients with melanoma. But the companies have also launched a phase III trial for lung cancer. And earlier this month BioNTech and Genentech announced that a personalized vaccine they developed in collaboration shows promise in pancreatic cancer, which has a notoriously poor survival rate.

Drug developers have been working for decades on vaccines to help the body’s immune system fight cancer, without much success. But promising results in the past year suggest that the strategy may be reaching a turning point. Will these therapies finally live up to their promise?

This week in The Checkup, let’s talk cancer vaccines. (And, you guessed it, mRNA.)

Long before companies leveraged mRNA to fight covid, they were developing mRNA vaccines to combat cancer. BioNTech delivered its first mRNA vaccines to people with treatment-resistant melanoma nearly a decade ago. But when the pandemic hit, development of mRNA vaccines jumped into warp drive. Now dozens of trials are underway to test whether these shots can transform cancer the way they did covid. 

Recent news has some experts cautiously optimistic. In December, Merck and Moderna announced results from an earlier trial that included 150 people with melanoma who had undergone surgery to have their cancer removed. Doctors administered nine doses of the vaccine over about six months, as well as  what’s known as an immune checkpoint inhibitor. After three years of follow-up, the combination had cut the risk of recurrence or death by almost half compared with the checkpoint inhibitor alone.

The new results reported by BioNTech and Genentech, from a small trial of 16 patients with pancreatic cancer, are equally exciting. After surgery to remove the cancer, the participants received immunotherapy, followed by the cancer vaccine and a standard chemotherapy regimen. Half of them responded to the vaccine, and three years after treatment, six of those people still had not had a recurrence of their cancer. The other two had relapsed. Of the eight participants who did not respond to the vaccine, seven had relapsed. Some of these patients might not have responded  because they lacked a spleen, which plays an important role in the immune system. The organ was removed as part of their cancer treatment. 

The hope is that the strategy will work in many different kinds of cancer. In addition to pancreatic cancer, BioNTech’s personalized vaccine is being tested in colorectal cancer, melanoma, and metastatic cancers.

The purpose of a cancer vaccine is to train the immune system to better recognize malignant cells, so it can destroy them. The immune system has the capacity to clear cancer cells if it can find them. But tumors are slippery. They can hide in plain sight and employ all sorts of tricks to evade our immune defenses. And cancer cells often look like the body’s own cells because, well, they are the body’s own cells.

There are differences between cancer cells and healthy cells, however. Cancer cells acquire mutations that help them grow and survive, and some of those mutations give rise to proteins that stud the surface of the cell—so-called neoantigens.

Personalized cancer vaccines like the ones Moderna and BioNTech are developing are tailored to each patient’s particular cancer. The researchers collect a piece of the patient’s tumor and a sample of healthy cells. They sequence these two samples and compare them in order to identify mutations that are specific to the tumor. Those mutations are then fed into an AI algorithm that selects those most likely to elicit an immune response. Together these neoantigens form a kind of police sketch of the tumor, a rough picture that helps the immune system recognize cancerous cells. 

“A lot of immunotherapies stimulate the immune response in a nonspecific way—that is, not directly against the cancer,” said Patrick Ott, director of the Center for Personal Cancer Vaccines at the Dana-Farber Cancer Institute, in a 2022 interview.  “Personalized cancer vaccines can direct the immune response to exactly where it needs to be.”

How many neoantigens do you need to create that sketch?  “We don’t really know what the magical number is,” says Michelle Brown, vice president of individualized neoantigen therapy at Moderna. Moderna’s vaccine has 34. “It comes down to what we could fit on the mRNA strand, and it gives us multiple shots to ensure that the immune system is stimulated in the right way,” she says. BioNTech is using 20.

The neoantigens are put on an mRNA strand and injected into the patient. From there, they are taken up by cells and translated into proteins, and those proteins are expressed on the cell’s surface, raising an immune response

mRNA isn’t the only way to teach the immune system to recognize neoantigens. Researchers are also delivering neoantigens as DNA, as peptides, or via immune cells or viral vectors. And many companies are working on “off the shelf” cancer vaccines that aren’t personalized, which would save time and expense. Out of about 400 ongoing clinical trials assessing cancer vaccines last fall, roughly 50 included personalized vaccines.

There’s no guarantee any of these strategies will pan out. Even if they do, success in one type of cancer doesn’t automatically mean success against all. Plenty of cancer therapies have shown enormous promise initially, only to fail when they’re moved into large clinical trials.

But the burst of renewed interest and activity around cancer vaccines is encouraging. And personalized vaccines might have a shot at succeeding where others have failed. The strategy makes sense for “a lot of different tumor types and a lot of different settings,” Brown says. “With this technology, we really have a lot of aspirations.”

Now read the rest of The Checkup

Read more from MIT Technology Review’s archive

mRNA vaccines transformed the pandemic. But they can do so much more. In this feature from 2023, Jessica Hamzelou covered the myriad other uses of these shots, including fighting cancer. 

This article from 2020 covers some of the background on BioNTech’s efforts to develop personalized cancer vaccines. Adam Piore had the story. 

Years before the pandemic, Emily Mullin wrote about early efforts to develop personalized cancer vaccines—the promise and the pitfalls. 

From around the web

Yes, there’s bird flu in the nation’s milk supply. About one in five samples had evidence of the H5N1 virus. But new testing by the FDA suggests that the virus is unable to replicate. Pasteurization works! (NYT)

Studies in which volunteers are deliberately infected with covid—so-called challenge trials—have been floated as a way to test drugs and vaccines, and even to learn more about the virus. But it turns out it’s tougher to infect people than you might think. (Nature)

When should women get their first mammogram to screen for breast cancer? It’s a matter of hot debate. In 2009, an expert panel raised the age from 40 to 50. This week they lowered it to 40 again in response to rising cancer rates among younger women. Women with an average risk of breast cancer should get screened every two years, the panel says. (NYT)

Wastewater surveillance helped us track covid. Why not H5N1? A team of researchers from New York argues it might be our best tool for monitoring the spread of this virus. (Stat)

Long read: This story looks at how AI could help us better understand how babies learn language, and focuses on the lab I covered in this story about an AI model trained on the sights and sounds experienced by a single baby. (NYT)