The study, published Jan. 30 in the Journal of Antimicrobial Chemotherapy, adds to growing evidence that lenacapavir may be a powerful new tool in the global anti-HIV drug arsenal. Approximately, 1.5 million people are living with HIV in Uganda.

“Our data shows that only 1.6% of the individuals studied are living with HIV strains that have any known lenacapavir-associated resistance mutations,” said senior author Dr. Guinevere Lee, assistant professor of virology in medicine at Weill Cornell Medicine. “That’s important because it shows lenacapavir is likely to be effective against strains of HIV circulating in East Africa.”

Since the 1990s, HIV drug combinations targeting different steps in the virus’ life cycle have been able to reduce virus load in patients to nearly undetectable levels. But drug resistance is a growing concern as the virus has evolved ways to thwart existing therapies. Lenacapavir, however, is the first drug to disrupt the protective capsid layer surrounding HIV’s genetic material (RNA), blocking the virus’s ability to reproduce and be transmitted from person to person.

Treatment twice a year with lenacapavir has been effective in patients who have never been treated and those with HIV strains that are resistant to other drugs. Last year, clinical trials showed that lenacapavir injections were 100% effective in preventing HIV infection among women in sub-Saharan Africa, who were HIV-negative.

However, little information was available about pre-existing resistance to lenacapavir in less well-studied HIV-1 strains like subtype A1 and D, which are more common in Eastern and Southern Africa. HIV-1 subtype B strains, which predominantly affect Europe and the United States, rarely have pre-existing mutations that would cause lenacapavir drug resistance.

Dr. Lee and her colleagues at Mbarara University of Science and Technology in Uganda and Massachusetts General Hospital in Boston helped fill that gap. They sequenced the capsid proteins from HIV-1 subtypes A1 and D from 546 Ugandan patients, who had never used antiretroviral therapy before. This approach allowed the investigators to examine naturally circulating viral variants.

They found that none of the patients had genetic mutations that would lead to major lenacapavir resistance. Only nine participants had minor lenacapavir resistance mutations that could partially reduce the effectiveness, but not enough to cause full resistance to the drug.

“Our study supports lenacapavir’s potential efficacy in this region. As lenacapavir is rolled out in East Africa, further studies will need to monitor for the emergence of drug-resistant strains,” Dr. Lee said. “It is important that we ensure HIV research reaches understudied communities where unique viral strains circulate.”

MIT researchers probed the inner workings of LLMs to better understand how they process such assorted data, and found evidence that they share some similarities with the human brain.

Neuroscientists believe the human brain has a “semantic hub” in the anterior temporal lobe that integrates semantic information from various modalities, like visual data and tactile inputs. This semantic hub is connected to modality-specific “spokes” that route information to the hub. The MIT researchers found that LLMs use a similar mechanism by abstractly processing data from diverse modalities in a central, generalized way. For instance, a model that has English as its dominant language would rely on English as a central medium to process inputs in Japanese or reason about arithmetic, computer code, etc. Furthermore, the researchers demonstrate that they can intervene in a model’s semantic hub by using text in the model’s dominant language to change its outputs, even when the model is processing data in other languages.

These findings could help scientists train future LLMs that are better able to handle diverse data.

“LLMs are big black boxes. They have achieved very impressive performance, but we have very little knowledge about their internal working mechanisms. I hope this can be an early step to better understand how they work so we can improve upon them and better control them when needed,” says Zhaofeng Wu, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this research.

His co-authors include Xinyan Velocity Yu, a graduate student at the University of Southern California (USC); Dani Yogatama, an associate professor at USC; Jiasen Lu, a research scientist at Apple; and senior author Yoon Kim, an assistant professor of EECS at MIT and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). The research will be presented at the International Conference on Learning Representations.

Integrating diverse data

The researchers based the new study upon prior work which hinted that English-centric LLMs use English to perform reasoning processes on various languages.

Wu and his collaborators expanded this idea, launching an in-depth study into the mechanisms LLMs use to process diverse data.

An LLM, which is composed of many interconnected layers, splits input text into words or sub-words called tokens. The model assigns a representation to each token, which enables it to explore the relationships between tokens and generate the next word in a sequence. In the case of images or audio, these tokens correspond to particular regions of an image or sections of an audio clip.

The researchers found that the model’s initial layers process data in its specific language or modality, like the modality-specific spokes in the human brain. Then, the LLM converts tokens into modality-agnostic representations as it reasons about them throughout its internal layers, akin to how the brain’s semantic hub integrates diverse information.

The model assigns similar representations to inputs with similar meanings, despite their data type, including images, audio, computer code, and arithmetic problems. Even though an image and its text caption are distinct data types, because they share the same meaning, the LLM would assign them similar representations.

For instance, an English-dominant LLM “thinks” about a Chinese-text input in English before generating an output in Chinese. The model has a similar reasoning tendency for non-text inputs like computer code, math problems, or even multimodal data.

To test this hypothesis, the researchers passed a pair of sentences with the same meaning but written in two different languages through the model. They measured how similar the model’s representations were for each sentence.

Then they conducted a second set of experiments where they fed an English-dominant model text in a different language, like Chinese, and measured how similar its internal representation was to English versus Chinese. The researchers conducted similar experiments for other data types.

They consistently found that the model’s representations were similar for sentences with similar meanings. In addition, across many data types, the tokens the model processed in its internal layers were more like English-centric tokens than the input data type.

“A lot of these input data types seem extremely different from language, so we were very surprised that we can probe out English-tokens when the model processes, for example, mathematic or coding expressions,” Wu says.

Leveraging the semantic hub

The researchers think LLMs may learn this semantic hub strategy during training because it is an economical way to process varied data.

“There are thousands of languages out there, but a lot of the knowledge is shared, like commonsense knowledge or factual knowledge. The model doesn’t need to duplicate that knowledge across languages,” Wu says.

The researchers also tried intervening in the model’s internal layers using English text when it was processing other languages. They found that they could predictably change the model outputs, even though those outputs were in other languages.

Scientists could leverage this phenomenon to encourage the model to share as much information as possible across diverse data types, potentially boosting efficiency.

But on the other hand, there could be concepts or knowledge that are not translatable across languages or data types, like culturally specific knowledge. Scientists might want LLMs to have some language-specific processing mechanisms in those cases.

“How do you maximally share whenever possible but also allow languages to have some language-specific processing mechanisms? That could be explored in future work on model architectures,” Wu says.

In addition, researchers could use these insights to improve multilingual models. Often, an English-dominant model that learns to speak another language will lose some of its accuracy in English. A better understanding of an LLM’s semantic hub could help researchers prevent this language interference, he says.

This research is funded, in part, by the MIT-IBM Watson AI Lab.

“Breast cancer incidence is increasing in U.S. women under 40, but until now, it was unknownif incidence trends varied by U.S. geographic region,” said Rebecca Kehm, PhD, assistant professor of Epidemiology at Columbia Mailman School, and first author. “Our findings can more accurately inform whether exposures that vary in prevalence across the U.S. also contributes to breast cancer risk in younger women.”

Using the U.S. Cancer Statistics database, the researchers analyzed age-adjusted breast cancer-incidence rates from 2001 to 2020 in women aged 25-39. They calculated the average annual percent change using statistical regression formulas and performed age-distribution analyses.

“Two-thirds of all cancers identified both in the U.S. and globally are diagnosed in women,” said Mary Beth Terry, PhD, professor of Epidemiology at Columbia Mailman School of Public Health, and senior author of the study.

From 2001 to 2020, breast cancer incidence in women under 40 increased by more than 0.50 percent per year in 21 states, while remaining stable or decreasing in the other states. Incidence was 32 percent higher in the five states with the highest rates compared to the five states with the lowest rates. The Western region had the highest rate of increase from 2001 to 2020; the Northeast had the highest absolute rate among women under 40 and experienced a significant increase over time The South was the only region where breast cancer under 40 did not increase from 2001 to 2020.

The overall incidence of early-onset breast cancer ranged from 28.6 per 100,000 in Wyoming to 41 cases per 100,000 people in Connecticut. The five states with the highest early-onset incidence from 2001 to 2020 were Maryland, New York, New Jersey, Hawaii, and Connecticut. Hispanic women had the lowest early-onset frequency rates in all regions, ranging from 26 per 100,000 in the Midwest to 32.6 per 100,000 in the Northeast.

Non-Hispanic White women were the only group to experience a statistically significant increase in early-onset breast cancer incidence across all four regions of the U.S. Non-Hispanic Black women had the highest incidence of early-onset breast cancer. This was true across the regions of the country.

The authors note the importance of investigating other risk factors including alcohol consumption, an established risk factor for breast cancer and which is known to vary across states and also be influenced by state alcohol policies.

“The increase in incidence we are seeing is alarming and cannot be explained by genetic factors, alone which evolve over much longer periods nor by changes in screening practices given that women under 40 years are below the recommended age for routine mammography screening,” noted Kehm.

“While the causes behind the rising incidence of early onset breast cancer are not yet fully understood, studying how trends vary across different population subgroups can offer valuable insights and help generate hypotheses for future research,” said Professor Terry. “We also are able to gain an understanding into the increase in breast cancer incidence among women who are not currently recommended for routine screening.”

Co-authors are Josephine Daaboul, Columbia Mailman School of Public Health and Fielding School of Public Health, University of California Los Angeles; and Parisa Tehranifar, Columbia Mailman School of Public Health.

The study was supported by the National Cancer Institute (R00CA263024).

A new study, led by UC San Francisco researchers, found that older, healthy volunteers, with lower concentrations of B12, but still in the normal range, showed signs of neurological and cognitive deficiency. These levels were associated with more damage to the brain’s white matter — the nerve fibers that enable communication between areas of the brain — and test scores associated with slower cognitive and visual processing speeds, compared to those with higher B12.

The study published in Annals of Neurology on Feb. 10.

The researchers led by senior author Ari J. Green, MD, of the UCSF Departments of Neurology and Ophthalmology and the Weill Institute for Neurosciences, said that the results raise questions about current B12 requirements and suggest the recommendations need updating.

“Previous studies that defined healthy amounts of B12 may have missed subtle functional manifestations of high or low levels that can affect people without causing overt symptoms,” said Green, noting that clear deficiencies of the vitamin are commonly associated with a type of anemia. “Revisiting the definition of B12 deficiency to incorporate functional biomarkers could lead to earlier intervention and prevention of cognitive decline.”

Lower B12 correlates with slower processing speeds, brain lesions

In the study, researchers enrolled 231 healthy participants without dementia or mild cognitive impairment, whose average age was 71. They were recruited through the Brain Aging Network for Cognitive Health (BrANCH) study at UCSF.

Their blood B12 amounts averaged 414.8 pmol/L, well above the U.S. minimum of 148 pmol/L. Adjusted for factors like age, sex, education and cardiovascular risks, researchers looked at the biologically active component of B12, which provides a more accurate measure of the amount of the vitamin that the body can utilize. In cognitive testing, participants with lower active B12 were found to have slower processing speed, relating to subtle cognitive decline. Its impact was amplified by older age. They also showed significant delays responding to visual stimuli, indicating slower visual processing speeds and general slower brain conductivity.

MRIs revealed a higher volume of lesions in the participants’ white matter, which may be associated with cognitive decline, dementia or stroke.

While the study volunteers were older adults, who may have a specific vulnerability to lower levels of B12, co-first author Alexandra Beaudry-Richard, MSc, said that these lower levels could “impact cognition to a greater extent than what we previously thought, and may affect a much larger proportion of the population than we realize.” Beaudry-Richard is currently completing her doctorate in research and medicine at the UCSF Department of Neurology and the Department of Microbiology and Immunology at the University of Ottawa.

“In addition to redefining B12 deficiency, clinicians should consider supplementation in older patients with neurological symptoms even if their levels are within normal limits,” she said. “Ultimately, we need to invest in more research about the underlying biology of B12 insufficiency, since it may be a preventable cause of cognitive decline.”

The experiments were designed to answer a question never before asked of wild fish: are they capable of telling people apart? Overall, little scientific evidence exists to show that fish can recognize humans at all. One captive-bred species, archerfish, was able to recognize computer-generated images of human faces in laboratory experiments. “But nobody has ever asked whether wild fish have the capacity, or indeed motivation, to recognize us when we enter their underwater world,” says Maëlan Tomasek, a doctoral student at MPI-AB and the University of Clermont Auvergne, France.

Now, a team from MPI-AB have asked; and the fish have responded. Wild fish can recognize individual humans. And, more than that, they follow specific divers they know will reward them. This finding, published in Biology Letters, lends credence to the possibility that fish can have differentiated relationships with specific humans.

The fish who volunteered

The research team conducted the study eight meters underwater at a research site in the Mediterranean Sea where populations of wild fish have become habituated to the presence of scientists. Their experiments took place in open water and fish participated in trials as “willing volunteers who could come and go as they pleased,” explains Katinka Soller, a bachelor student from MPI-AB who was co-first author on the study with Tomasek.

The first experimental phase — the training — tested if fish could learn to follow an individual diver. The training diver, Soller, started by trying to attract the attention of local fish; she wore a bright red vest and fed fish while swimming a length of 50 meters. Over time, Soller removed the conspicuous cues until she wore plain dive gear, kept the food hidden, and fed fish only after they had followed her the full 50 meters.

Of dozens of fish species inhabiting the marine station, two species of seabream in particular willingly engaged in the training sessions. Sea bream are best known to us as fish that we buy to eat, yet they surprised the scientists by their curiosity and willingness to learn.

“Once I entered the water, it was a matter of seconds before I would see them swimming towards me, seemingly coming out of nowhere,” says Soller. Not only were bream learning to follow her, but the same individuals were showing up day after day to join the lessons. Soller even took to giving them names: “There was Bernie with two shiny silver scales on the back and Alfie who had a nip out of the tail fin,” she says.

After 12 days of training, roughly 20 fish were reliably following Soller on training swims and she could recognize several of them from physical traits. By identifying individual fish participating in the experiment, the stage was set for the next experimental phase: testing if these same fish could tell Soller apart from another diver.

The two-diver test

This time Soller dived with Tomasek whose dive gear differed slightly from hers, notably in some colorful parts of the wetsuit and fins. Both divers started at the same point and then swam in different directions. On the first day, the fish followed both divers equally. “You could see them struggling to decide who to chase,” says Soller.

But Tomasek never fed the fish who followed him, so from the second day, the number of fish following Soller increased significantly. To confirm that fish were learning to recognize the correct diver, the researchers focused on six fish out of the large group to study individually, finding that four of these showed strong positive learning curves over the experiment. “This is a cool result because it shows that fish were not simply following Katinka out of habit or because other fish were there,” says Tomasek. “They were conscious of both divers, testing each one and learning that Katinka produced the reward at the end of the swim.”

But when Soller and Tomasek repeated the trials, this time wearing identical diving gear, the fish were unable to discriminate them. For the scientists, this was strong evidence that fish had associated the differences in the dive gear, most likely the colors, with each diver. “Almost all fish have color vision, so it is not surprising that the sea bream learned to associate the correct diver based on patches of color on the body,” says Tomasek.

Fish know how we look

Underwater, we do the same. “Faces are distorted by diving masks, so we usually rely on differences between wetsuits, fins, or other parts of the gear to recognize each other,” says Soller. With more time, the authors say, fish might have learned to pay attention to subtler human features, like hair or hands, to distinguish divers. “We already observed them approaching our faces and scrutinizing our bodies,” adds Soller. “It was like they were studying us, not the other way around.”

This study corroborates many anecdotal reports of animals, including fish, recognizing humans; but it goes further by performing dedicated experiments in completely natural contexts. Finding that wild fish can quickly learn to use specific cues to recognize individual human divers, it stands to reason that many other fish species, our pets included, can recognize certain patterns to identify us, the scientists say. This mechanism is the foundation for special interactions between individuals, even across species.

Senior author Alex Jordan, who leads a group at MPI-AB, says: “It doesn’t come a shock to me that these animals, which navigate a complex world and interact with myriad different species every minute, can recognize humans based on visual cues. I suppose the most surprising thing is that we would be surprised they can. It suggests we might underestimate the capacities of our underwater cousins.”

Adds Tomasek: “It might be strange to think about humans sharing a bond with an animal like a fish that sits so far from us on the evolutionary tree, that we don’t intuitively understand. But human-animal relationships can overcome millions of years of evolutionary distance if we bother to pay attention. Now we know that they see us, it’s time for us to see them.”

SACs are one of the most crucial catalysts. They play a pivotal role in enhancing efficiency in diverse applications including chemical industries, energy conversion, and environmental processes. For water purification in particular, SACs can overcome the limitations of traditional heterogeneous catalysts such as the kinetics, catalytic selectivity, and stability — paving a promising way for the advancement of efficient and sustainable water purification technologies.

However, the development of SACs frequently employs time-consuming trial-and-error methods, and the typical synthesis methods often lack a high level of control. To avoid a process that essentially involves taking shots in the dark, researchers took a data-driven approach where they rapidly and accurately predicted which SACs would have the best performance before even starting to make them. They compared 43 metals-N₄ structures comprising transition and main group metal elements using a hard-template method.

Following this strategy, they determined that the best candidate was a well-designed Fe-SAC with a high loading of Fe-pyridine-N4 sites (~3.83 wt%) and highly mesoporous structure. It successfully exhibited ultra-high decontamination performance (rate constant of 100.97 min^-1 g^-2).

“The optimized Fe-SAC can also continuously operate for 100 hours,” remarks Associate Professor Hao Li of WPI-AIMR, “To our knowledge, this represents one of the best performances of wastewater purification on Fenton-like catalysts — which are reagents used for water purification — reported so far.”

Density functional theory calculations revealed that the underlying mechanism was that the SAC reduced the energy barrier of the rate-determining step, which is intermediate O* formation. This resulted in the highly selective generation of singlet oxygen, which has been shown to break down pollutants to help purify water.

To make sure the data-driven prediction had accurately selected this “best” candidate, the research team looked at five other metals-N4 structures (i.e., Fe, Co, Ni, Cu, and Mn) with different theoretical activities. They confirmed that Fe-SAC truly exhibited the most excellent Fenton-like performance among the five selected SACs, agreeing well with the data-driven prediction.

The close integration of a data-driven method with a precise synthesis strategy provides a novel paradigm for the rapid development of high-performance catalysts for environmental fields, and other fields that involve sustainable energy and catalysis. Moving forward, they aim to develop an efficient and user-friendly workflow for the rapid and effective design of catalysts.

Those interested in incorporating the method into their own work can view the experimental data and computational structures in the Digital Catalysis Platform (DigCat): the largest experimental catalysis database reported to date, developed by the Hao Li Lab. The findings were also published in Angewandte Chemie International Edition on January 31, 2025.

The article processing charge (APC) was supported by the Tohoku University Support Program.