In 1996, cryobiologist Gregory Fahy walked into his physician’s office and talked his way into a month’s supply of growth hormone. His hope, bolstered by a single study in rats1, was that the injections would help him to regrow his thymus — a peculiar immune organ that atrophies and practically disappears as people age. Regenerating it, Fahy thought, would help him to live a longer, healthier life. Tech titans are hacking their bodies for a longer life: is there science behind their methods? The improvement was obvious, at least according to magnetic resonance imaging (MRI) scans, Fahy says. His functional thymic mass nearly doubled2. Whether it made him feel any younger, however, was less clear. “I was only 46 at the time, and more or less in peak health,” he says, although subsequent attempts to regrow the tissue left him feeling “energized and invincible”. What started as an unregulated self-experiment has developed into a series of small clinical trials run by a biopharmaceutical company called Intervene Immune in Torrance, California, at which Fahy is chief scientific officer. The company is not alone. An explosion of thymus research has taken place over the past three years, stimulated by reports in the literature that the health of the organ, once thought dispensable, is a probable indicator of a person’s overall health. Excitement intensified when a pair of studies published this year reported that dwindling thymic health correlates with an increased risk of death3,4. Investors are responding. In January, a biotechnology company in Basel, Switzerland, called TECregen raised 10 million Swiss francs (US$12.4 million) to develop therapies to regenerate the thymus, which might help to slow ageing and prevent cancer, according to the company’s website. It has not disclosed a drug candidate. Last October, Zag Bio, a thymus-focused biotech firm in Cambridge, Massachusetts, launched with $80 million in funding. Venture-capital and pharmaceutical companies have been reaching out to Fahy and other scientists and starting their own research programmes. “The interest now in this field is enormous,” says Marcel van den Brink, chief physician and president of City of Hope, a cancer hospital and research centre in Duarte, California, who has studied the thymus since 1996. There was a “huge perception that I was wasting my time”, he says. But now, “people are contacting us” and taking the field seriously. (van den Brink is a scientific co-founder of a biotech company called Thymofox in Cambridge, Massachusetts.) Forgotten organ For decades, people thought that the thymus was vestigial. It has been “forgotten” and “disrespected”, says Georg Holländer, an immunologist at the University of Oxford, UK, who is a co-founder of TECregen. In the 1920s, some researchers thought that the organ was involved in producing eggshells in birds but had limited roles in mammals. Decades later, Nobel-prizewinning biologist Peter Medawar referred to the thymus as “an evolutionary accident of no very great significance”. How your brain controls ageing — and why zombie cells could be key They were wrong. The organ’s importance had escaped scientists for so long in part because mammals that have their thymuses surgically removed seem to be perfectly healthy. It wasn’t until the 1960s, when immunologist Jacques Miller performed the surgeries on newborn mice, that the function of the organ was discovered5. Infant mice without a thymus often succumbed to infection and died. They had few immune cells in their lymph nodes. Miller found that the thymus creates T cells, a cell type that he also identified around the same time and which is named after the organ. T cells, we now know, are a crucial part of the immune system that attacks cancer cells and fights infections. In humans, the thymus shrinks drastically over a person’s lifetime. After puberty, most of its cells turn into non-functional fatty tissue. The production of T cells slows. At age 40, the thymus produces roughly one-quarter of the T cells it did at age 8. At 65, it makes just 10%6. Focus on function For this reason, many scientists assumed that fresh T cells were no longer needed for older individuals. “People are under the impression that it’s only important to make your T cells when you’re young,” says Jennifer Cowan, an immunologist at University College London. “And that’s just quite clearly not the case.” Three discoveries changed that paradigm. First, a 2023 study found that people who had their thymus surgically removed were almost three times more likely to die and twice as likely to develop a cancer five years after the surgery than were people who had not had their thymus removed7. “It was very striking,” Cowan says, “to the point where people aren’t going to want to have their thymus removed if they’ve read that paper.” Then, in March, two back-to-back studies collectively analysed data from more than 31,000 individuals and extended the findings to people with intact thymuses that had naturally withered with age. One found that a smaller thymus correlates with a higher risk of death as well as a higher risk of developing cancer and cardiovascular diseases such as heart failure3. The second found that a smaller thymus is linked to worse survival after receiving immunotherapy for cancer4. Some scientists point out that the studies only offer correlations. “When you’re looking retrospectively, there’s a lot of pieces of the puzzle that are just not knowable,” says Daniel Boffa, a thoracic surgeon at Yale University in New Haven, Connecticut. He and his colleagues performed a study that found little difference in the risk for recurring tumours among people who had all or part of their thymus removed to treat thymic cancer8. Nevertheless, commercial and academic interest in the thymus has surged. “Regrowth has become a big issue,” says Yousuke Takahama, a developmental biologist at the US National Cancer Institute in Bethesda, Maryland. “Just like rebooting your computer, it’s like rebooting our immune system.” Testosterone therapy is trending. Who really needs it, and why? Scientists think that the thymus has a large regenerative potential. For example, the thymus withers during pregnancy and then recovers following delivery, says David Scadden, a haematologist at the Massachusetts General Hospital in Boston, who co-founded a company called Lightning Biotherapeutics in Cambridge, Massachusetts. But there are few proven methods for regenerating the thymus. Androgen deprivation therapy, a standard treatment for prostate cancer, can increase T-cell output from the thymus9. So can calorie restriction10. A treatment for children born without a thymus received regulatory approval in 2021, but it is costly and complex, involving transplantation of carefully processed donor tissue into a child’s thigh muscle. As for a treatment to regrow an ageing thymus, the approach that is now furthest along is Fahy’s cocktail of growth hormone and other compounds, injected four times a week. A group of ten volunteers showed signs of reversed ageing after one year on the drugs. Fahy’s team looked at the participants’ ‘epigenetic clocks’ — a collection of chemical modifications to chromosomes that reliably shifts with age. Results suggested that treatment turned back people’s epigenetic clocks by an average of two and a half years11. Fahy says he predicts that people will take the injections for a year and then take a break for four years, and repeat the cycle. Further trials will test whether his therapy could be tweaked to minimize side effects and improve metrics such as muscle strength, body composition and white-matter mass in the brain. “It’ll be a couple of years before we have results of those studies,” he says. Researchers are also studying other treatments that they hope will avoid the potential increased cancer risk and blood-sugar spikes that can come with taking growth hormones. Is the peptide craze backed by science? The promise behind the hype Francisco Leon, chief executive of the biotech company Tolerance Bio in Philadelphia, Pennsylvania, for example, says that his company has an antibody that slows down the shrinking of the thymus. The company has not disclosed what the antibody targets, but it plans to test it in clinical trials in 2027. “This could be a longevity drug,” Leon says. “We are going to be able to prevent and treat diseases of ageing.” Thymmune Therapeutics, a biotech firm based in Cambridge, is developing off-the-shelf stem cells that they hope can grow into thymus tissue when injected into a person. Several challenges remain. The biggest barrier is targeting drugs to the thymus specifically — most of the organ is composed of cells that lack a unique surface protein. And many candidate molecules are delivered intravenously, which is inconvenient for a long-term treatment. Moreover, researchers say that most studies today do not look at the mechanisms of how T cells produced by a ‘regenerated thymus’ might improve health. “It’s easy to see that the size is restored, but it’s less clear whether or not the full functionality is there,” says Ann Griffith, a microbiologist at the University of Texas Health Science Center at San Antonio who is on the advisory board of a company in this space. It’s unclear when a therapy might be available more widely, but it will probably take years. “We don’t fully understand how the interventions really regenerate the tissue and we’re really scratching the surface,” Griffith says. “We’re at the very beginning of the process.”

The Maya temple Tikal in Guatemala is about one day's walk from Xultun, where researchers discovered mathematical formula scribbled on the walls.Credit: Kryssia Campos/Getty A mathematical formula inscribed on a wall at the Maya site of Xultun in Guatemala has revealed the name of an important Maya mathematician-astronomer for the first time. Researchers suggest Sak Tahn Waax, or ‘White-Chested Fox’, was a scholar comparable with mathematical giants of the past. In a study published 14 July in the journal Antiquity1, Heather Hurst, an archaeologist at Skidmore College in Saratoga Springs, New York, and her colleagues describe their analysis of a mathematical text from a chamber in Xultun that was originally excavated in 20112. The chamber’s walls are painted with human figures and hieroglyphic texts. These include mathematical calculations based on astronomical calendars, which were used by the Maya people to decide the timing of events such as the inaugurations of kings. Hurst and her colleagues suggest that the chamber was a workspace for scribes making codices in the mid-eighth century ad. The authors analysed one set of hieroglyphs in particular, referred to as Text 19. Hurst says that this set of mathematical calculations expresses the relationships between several calendar systems in a playful manner that hasn’t been seen before in Mayan texts. “I think it was a mathematical flex. Somebody was saying ‘I’ve got this amazing pattern, and it’s so good it needs to be written down’. It was like, ‘Boom! Mic drop!’,” says Hurst. “The discovery shows people that the Maya were very clever, creative, intellectually curious people who taught and learnt and sometimes did math for the sake of it,” says Eric Heller, an archaeologist at the University of Southern California Dornsife. Identity revealed Text 19 is a small, L-shaped group of eleven hieroglyphs with a combined height of about 10 centimetres. Hurst and her colleagues found that the first nine hieroglyphs of the set encode the Maya calendar and astronomical cycles. The formula shows how one 2,920-day cycle could be divided up into the calendar units used by the Maya people. This 2,920-day cycle was important because it tied together key astronomical cycles, corresponding to both five Venus cycles (584 days each) and eight solar years (365 days each). However, the Text 19 calculations also relate the 2,920 days to Uinal (months with 20 days), Tzolkin (the 260-day sacred calendar), Tun (a year with 360 days) and Mars years of 780 days. “It’s just super nerdy math,” says Hurst. The hieroglyphs also show only partial dates, which made them hard to decipher. “They’re doing this abbreviated shorthand, so they give you the first half of a notation and the second half is implied.” The mathematical formula on Text 19 appear as glyphs.Credit: Photograph by G. Ware, courtesy of the San Bartolo-Xultun Regional Archaeological Project Until now, the identities of the mathematician-astronomers behind such calculations had remained mysterious. Hurst and her colleagues found a phrase in the penultimate hieroglyph in Text 19 meaning “so says”. This was followed by the name Sak Tahn Waax, in the final hieroglyph, suggesting that the writer was taking or giving credit for the calculation. “We know it’s a male name because it’s missing a prefix,” she says. The fact that the writer is named is significant, because it suggests that mathematicians were recognized in Maya society as much as artists were, says Gerardo Aldana, an anthropologist at the University of California, Santa Barbara. Equalling the greats Hurst and her colleagues say the work also shows that Sak Tahn Waax should be considered on the same level as some of the great astronomer-mathematicians of history, such as Archimedes, Claudius Ptolemy and Muḥammad ibn Mūsā al-Khwārizmī. Ancient DNA from Maya ruins tells story of ritual human sacrifices Aldana agrees that the mathematical abilities shown here and in Maya manuscripts such as the Dresden codex, which dates back to the eleventh or twelfth century, are impressive when compared with maths from contemporaneous civilizations. The Maya people grasped the likes of positional notation, advanced arithmetical operations, algebraic relationships, negative numbers and multiplicative factors — but we don’t have any evidence that they made the leap to geometry, he says. “They’re not picking up that there’s a geometry you can model to predict eclipses much more accurately than you can if you’re just relying on algebra,” says Aldana, adding that this could be down to the fact that Mesoamerica was isolated from the rest of the world at the time. The Greeks got geometry from the Egyptians, the Islamic world and India got their maths from the Greeks and Europe got its maths from Islam, he says. “You’re seeing all of these different cultures interacting and passing on ideas, but in Mesoamerica you only have one culture, and so I think that’s maybe why it doesn’t go in the same direction.”

The bottom of the ocean has been mapped in much less detail than has the surface of Mars.Credit: US Geological Survey/Science Photo Library The ocean’s depths are among the ultimate frontiers for scientific exploration. Scientists have visually surveyed less than 0.001% of the sea floor, yet oceanography is essential to understanding the climate, ecosystems and processes of deep Earth — including earthquakes and the resulting tsunamis. Now, emerging technologies, including advanced drilling technologies, seismic sensors and methods to convert deep-sea cables into a giant seismic-surveillance network, are poised to crack open these mysteries. Nature has looked into how some of these tools work, and the geoscience questions that they could help to answer. Mapping the mantle Plate tectonics is driven by convection, or the churning of the mantle — the layer of mostly solid rock under Earth’s crust that makes up more than 80% of the planet’s volume. But the mantle’s inner geological workings remain mostly a mystery. Researchers have begun to map the process of convection by measuring how different rock densities affect the propagation of seismic waves — using ocean-bottom seismometers (OBSs) that can run on batteries for one year or more. OBS studies have helped to show that the motion of rock in the mantle is “like a lava lamp”, says Ana Ferreira, a seismologist at University College London (UCL). “Imagine a pan with boiling syrup, but different types of syrup of different densities,” she says. Particularly hot ‘plumes’ (made of solid mantle rock that’s hotter than the surrounding mantle) create mid-ocean chains of volcanoes, such as those in Iceland or the Hawaiian archipelago. Researchers have used ocean-bottom seismometers, shown here being lowered into the water, to study Earth’s mantle. Credit: NASA Image Collection/Alamy Pioneering research in the 1990s focused on the Pacific Ocean, but scientists are now expanding their studies to the rest of the oceans. One such project, called UPFLOW, is currently analysing the data from OBSs that Ferreira and her collaborators deployed in the Atlantic around the Azores, Canary Islands and Madeira. In a separate project, the team showed how ocean-bottom tools could aid scientists during a crisis. A storm of earthquakes shook the volcanic São Jorge Island in the Azores in March 2022, causing fears of an upcoming eruption. Ferreira and her collaborators rushed to set up six OBSs around the island and mapped the magma activity under the volcano. In the end, there was no eruption. “The magma reached to within a kilometre of the surface. Then it stalled,” says UCL seismologist Stephen Hicks, a co-author of the study1. Drilling deep into the mantle Ultimately, scientists would like to understand the depths of Earth. One grand challenge is to drill through the crust’s lower boundary — called the Mohorovičić discontinuity — and take the first-ever pristine samples of the underlying mantle. Researchers have high expectations that a new oceanographic research ship — China’s Meng Xiang, which is Mandarin for ‘dream’, launched in late 2024 — could for the first time collect samples directly from the mantle. Meng Xiang is equipped to drill as far down as 11,000 metres below the sea surface — deeper than any scientific vessel so far. Peter Bijl, a palaeo-oceanographer at Utrecht University in the Netherlands, was invited to visit the ship during a workshop in Guangzhou, China, in late 2024 and was impressed by the array of on-board laboratories and facilities. “It had everything one could possibly need on a ship, and more,” Bijl says. Researchers plan to use the deep-sea drilling vessel Meng Xiang to take samples of Earth’s mantle. Credit: Chen Chuhong/China News Service/VCG via Getty Detecting tsunamis with hydrophones Tsunamis can start at one end of an ocean and end up causing widespread destruction at the other. Usama Kadri, an applied mathematician at Cardiff University, UK, and his collaborators have developed an ocean-based technique to predict whether and where a tsunami will cause destruction, up to several hours before its waves hit land. Their method relies on hydrophones, or underwater microphones, that can detect low-frequency sounds that are produced at the point of origin of a tsunami. Kadri and his collaborators created a mathematical model that uses data from hydrophones to simulate how tsunami waves propagate around the world2. “The strength of the software is that it can make calculations globally in less than 30 seconds,” Kadri says. Crucially, the underwater sounds travel three times faster than does the tsunami itself, so they can reach the hydrophones hours before the tsunami does. This could give local authorities precious time to evacuate local people from low-lying coastal areas. Deep-sea cables Researchers have found a way to detect seismic events in sea-floor crusts — through hundreds of thousands of kilometres of underwater optical fibres that are the arteries of the Internet. Geophysicists at the Helmholtz Centre for Geosciences in Potsdam, Germany, and their colleagues first showed in 2018 that existing networks could be used to detect seismic waves3. The technique, which was originally developed for detecting submarines and was called distributed acoustic sensing (DAS), exploits tiny imperfections that exist every few metres in the glass of an optical fibre. If the cable is stretched by seismic waves travelling along its length, light shining down the cable will be reflected off the imperfections with slightly different timing from usual. This enables a DAS device onshore to track a seismic wave as it propagates along the sea floor. “It’s kind of amazing that it works,” says Vala Hjörleifsdóttir, a seismologist at Reykjavik University. Ocean floor witnessed splitting apart for the first time — releasing lava DAS “took the Earth-science community by storm”, says Mikael Mazur, an optical-communications researcher at Nokia Bell Labs in Summit, New Jersey. But it works only near coasts, because optical fibre connections are interrupted every 100 kilometres or so by undersea ‘repeaters’ that amplify the signals. Mazur and his collaborators have now proposed4 a way to extend the idea, such that it works over thousands of kilometres. They use existing optical bypasses that were built in the repeaters that help to locate where an undersea cable is damaged. The team tested the technique by detecting the waves from the 8.8-magnitude earthquake that shook eastern Russia’s Kamchatka peninsula last year, using a 4,400-kilometre fibre that connects Hawaii to California. Although they could not attain the metre-by-metre resolution of the original DAS, they showed that they can track the motion of seismic waves with a resolution of 100 metres, without the need for any extra undersea equipment.

Researchers found a four-carbon sugar called erythrulose in a cloud of gas and dust at the centre of the Milky Way (image from Spitzer Space Telescope shown here).Credit: NASA, Caltech, Susan Stolovy (SSC, Caltech) Call it an extra-sweet discovery. A group of astronomers has detected a sugar molecule swirling inside a cloud of gas and dust near the centre of our galaxy. They are calling the molecule — a compound with four carbon atoms called erythrulose — the first true sugar spotted in ‘interstellar’ space. The findings, reported today in the journal Nature Astronomy1, could help clarify how life on Earth began. Asteroid fragments upend theory of how life on Earth bloomed “This is an incredibly exciting result,” says Brett McGuire, an astrochemist at the Massachusetts Institute of Technology in Cambridge. “Astronomers have, for a very long time, been pushing to detect sugars in space.” That’s because they’ve already seen hints that sugars on Earth that are essential to life originated in outer space. For instance, the five-carbon sugar ribose was previously found in some multi-billion-year-old meteorite samples2,3, suggesting that space rocks might have smashed into Earth and delivered this and other sugars. In 2000, astronomers reported the detection of the two-carbon molecule glycolaldehyde — sometimes considered the simplest sugar molecule — in the interstellar space between stars4. But although glycolaldehyde can act similarly to a sugar, “it is not formally a sugar”, McGuire says. True sugars, he adds, must have a backbone of at least three carbon atoms. Astronomers have continued to scan space for these bigger molecules ever since. A surprise signature Izaskun Jiménez-Serra, an astronomer at the Spanish National Research Council in Madrid, and her colleagues were among this group, but hadn’t had any luck while scouring the night sky. Then, in 2022, Emilio Cocinero, a physical chemist at the University of the Basque Country in Leioa, Spain, offered to share spectroscopic data for erythrulose — that is, the specific fingerprint of wavelengths that the molecule emits. Erythrulose has four carbons (grey). Red atoms are oxygens, and white atoms are hydrogens.Source: Ref. 1 “I said, ‘Okay, why don’t you send me the information, and then I’ll check whether we see it in our data’,” Jiménez-Serra says. She was sceptical at first. But to her surprise, the erythrulose fingerprint showed up in the team’s observations of a molecular cloud near the centre of the Milky Way. And using Spain’s Yebes 40-metre and IRAM 30-metre radio telescopes for deeper observations, Jiménez-Serra and her team saw the signal even more clearly. Feedstock for life? Molecular clouds are the birthplace of stars and planets. The team proposes that if sugar exists in these molecular clouds, it could be transferred to bodies such as asteroids and comets as a star system forms. According to one theory, Earth underwent a period of heavy bombardment from asteroids and comets around four billion years ago. Such bodies might have brought sugar molecules to the planet’s surface, Jiménez-Serra says. JWST spots the most distant ‘smoke’ molecules ever seen in space The results are especially tantalizing because erythrulose could have “provided the feedstock for the first nucleic acids” — simpler versions of building blocks of modern-day DNA and RNA, she adds. “That’s why the detection of erythrulose is so relevant for the origins of life.” The study has made scientists “confident” that these molecules have been detected, says Anthony Remijan, an astrophysicist at the Green Bank Observatory in West Virginia. At the same time, he adds, “I still haven’t wrapped my head around the fact that we have this four-carbon sugar that’s been detected, and yet the three-carbon sugars still elude us.” As for future directions, Jiménez-Serra is curious to see whether even-larger sugars can build up in interstellar space. In particular, the five-carbon ribose that is the basis of modern-day nucleic acids “would be an interesting molecule to detect”, Remijan says. “An actual building block of RNA and DNA — that would be the next big thing.”

Researchers are trying to make human sperm (shown here in a colourized electron microscope image) from stem cells in the lab.Credit: Juergen Berger/Science Photo Library It sounds like science fiction: collecting a person’s blood cells, engineering them so they eventually transform into immature sperm, and then incubating them in a tiny pouch grown on a mouse’s kidney. The quest to make babies with lab-grown eggs and sperm But it’s not. Today, a team of researchers reported in the journal Cell Stem Cell1 that it has successfully carried out the procedure, with the ultimate goal of making mature human sperm in the laboratory. For now, that goal remains elusive. The lab-grown cells stopped developing at an immature stage. Many hurdles need to be overcome to create mature sperm in the lab, says Eoin Whelan, a reproductive biologist at the University of Pennsylvania in Philadelphia who is a member of the research team. But the latest feat is a step in the right direction. In the meantime, the procedure could be used to study early stages of human sperm development and to hunt for reasons behind male infertility. Around 40% of male-infertility cases have no known cause. “We are approaching this from a basic science perspective,” Whelan says. “We are a long way from clinical application.” An elusive goal Some of the potential clinical applications are controversial, particularly the idea of using lab-grown sperm or eggs to make babies. Although some researchers hope that the approach could be used to treat specific cases of infertility, the practice also raises ethical concerns. One of these is that the technique could make it easier to genetically modify reproductive cells to produce ‘designer babies’. Making mice with two dads: this biologist rewrote the rules on sexual reproduction So far, a handful of researchers have been able to make mouse eggs and sperm from mouse skin cells. These cells are genetically reprogrammed so that they transform into induced pluripotent stem (iPS) cells — ‘youthful’ cells that act like those in embryos and can be chemically or genetically coaxed to turn into completely different cells, such as sperm. One team even used the approach to generate offspring from two male mice2. But researchers have not been able to translate these successes in mice to humans or other primates because of differences in how the species develop. It is also difficult to study human fetal development, so researchers aiming to understand and recapitulate the earliest days of sperm and egg development are left fumbling in the dark. “In humans, the work is so behind,” says Kotaro Sasaki, a developmental biologist also at the University of Pennsylvania. Step by step Sasaki and his colleagues previously determined how to convert human iPS cells into those that resemble early, embryonic cells that eventually give rise to eggs and sperm3. The researchers then mixed those immature cells with non-reproductive cells found in the testicles of developing mice4. The non-reproductive mouse cells can provide the protection and nutrients needed to support sperm development. This allowed Sasaki and his colleagues to tiptoe closer to growing mature sperm, but the cells did not progress beyond the early stage of sperm development that is found in human fetuses. Researchers mixed non-reproductive cells from mouse testicles (cyan) with immature macaque sperm cells (red and green) and transplanted them into a pouch on a mouse kidney. A month after transplantation, the mouse cells had arranged themselves into tubular structures like those in testicles (shown in this image).Credit: Whelan, E.C. et al./Cell Stem Cell Now, Sasaki, Whelan and their colleagues have taken the process further by transplanting their mixture of cells into living mice — on a region of the kidney that is known to be well-suited for supporting transplanted tissues. Once in that pouch, the transplanted cells self-organized into tubular structures similar to those in testicles where sperm are produced. Six months after the transplant, the human cells had developed into spermatogonia, a type of cell that can eventually give rise to mature sperm. Gene activity in these immature sperm cells closely resembled that seen in normal human spermatogonia. But most of the human cells halted their development at this stage, and none went on to form mature sperm. The same was true when the team performed the experiments in macaque monkeys. Maturity blockers? Sasaki says that species differences between the developing human spermatogonia and the supporting mouse cells could be to blame. Sperm development is complex, requiring input from non-reproductive cells in the surrounding tissue, as well as hormonal signals from distant organs. These factors might not be identical in mice and humans. People are having fewer babies: is it really the end of the world? Incubating the human cells with non-reproductive cells that are also from humans could help the cells to develop into mature sperm, says Amander Clark, a developmental biologist at the University of California, Los Angeles. “That’s the technological innovation that’s going to come next,” she says. Another important next step will be to confirm whether the spermatogonia are functional. In humans, researchers cannot do so without running afoul of ethical concerns, so Sasaki hopes to do these experiments in macaques. “At the end of the day, you’ve got to make sperm — and a baby,” says Kyle Orwig, a reproductive biologist at the University of Pittsburgh in Pennsylvania. “That’ll demonstrate that you’ve got bona fide cells.” In the meantime, several private companies are also working to generate lab-grown eggs and sperm. Last month, a biotechnology firm called Conception in Berkeley, California, reported that it had grown immature human eggs from stem cells in the lab. In May, another company, Paterna Biosciences in Salt Lake City, Utah, announced that it had grown mature sperm. But Paterna starts its process using immature sperm cells collected from testicles, rather than attempting to recapitulate the earliest stages of sperm development using iPS cells. Regardless, such claims are difficult to evaluate, Sasaki says. “They don’t show any data,” he says. “If you want to make a big claim, you need huge evidence.”

Posting studies on preprint servers such as bioRxiv is common practice in many scientific fields.Credit: Michael Szebor/Nature The central conclusions of biomedical preprints rarely change following peer review in a journal1, according to a study posted on the preprint server bioRxiv this month. The research also found that studies that appeared first as preprints are retracted at roughly half the rate of papers that did not appear online before being in a peer-reviewed journal. The authors say the findings suggests that preprints are a reliable source of information, although some scientists say the finding should be interpreted more cautiously. Posting preprints is common practice in science these days, but Ruslan Rust, a neuroscientist at the University of Southern California in Los Angeles, says that he often hears fellow scientists say that they are unreliable. In his experience, peer review does not typically lead to major changes in the contents of a study. Rust wanted to see if this held true across different fields of biomedical research. Using a large language model (LLM), Rust and his colleague extracted the main scientific conclusion from the abstracts of 72,644 biomedical manuscripts that were first uploaded to bioRxiv between 2018 and 2025. The model then assessed how much the abstracts had changed compared with their eventual peer-reviewed versions. The study, which has not been peer reviewed, reports that 39.9% of main conclusions were unchanged between the preprint and journal-published abstract, another 50% underwent only minor revisions. Slightly more than 10% went through major changes. When conclusions did change, they were more likely to become more cautious than more confident after peer review, the study found. Around 8.4% of main findings adopted more cautious language after peer review, whereas 4.2% used more confident wording. The extent of revision also varied across disciplines. Major changes occurred in only 7.2% of bioinformatics papers compared with 17.5% of microbiology studies. The authors also found that the frequency of major revisions declined over time, reducing from 17% among papers posted in 2019 to 5.7% in 2024. Julian Sienkiewicz, who studies artificial intelligence tools and data exploration at the Warsaw University of Technology, says the decrease in major revisions over time could indicate that peer reviewers are overloaded and might not be reading papers thoroughly. Rust suggests that the decrease reflects a change in how people use preprints. In earlier years after bioRxiv was launched, particularly during the COVID-19 pandemic, scientists were under pressure to post their findings in a short time. This meant many papers had to undergo major revisions before publication, he added. In the past few years, some manuscripts might have already included reviewer feedback in the first preprint version that was posted online, he adds. Fewer retractions Rust and his colleague also found that papers that first appeared as preprints were retracted at a rate of 8.1 per 10,000 papers, compared with 18.7 per 10,000 comparable papers that had never been posted as preprints. The authors caution that the comparison is observational, based on relatively few retractions and does not prove that posting a paper as a preprint reduces the likelihood of a retraction. Reacting to the findings on the professional-networking platform LinkedIn, some researchers noted that preprints are subject to strong selection bias depending on who posts them and which studies are posted. Rust agrees that differences between the types of author who choose to post preprints and those who do not could contribute to the pattern his team found. “I think those who post preprints are open about sharing data,” says Rust. “In general, there is evidence that sharing raw data leads to better and more reproducible science,” he adds. The study only included preprints that were eventually published in a journal and so did not assess the veracity of unpublished preprints. Rust acknowledges this is a limitation of the analysis but adds that previous estimates suggest that roughly two-thirds of preprints posted before 2017 on bioRxiv are eventually published in peer-reviewed journals2. “For most scientists, reputation is important, and they would not upload something they would not be comfortable with publishing in a journal later.” Sienkiewicz also says that LLMs do not always accurately match preprints to their corresponding peer-reviewed versions. Sienkiewicz was part of the team that created PreprintToPaper, a data set that connects bioRxiv preprints to published papers3. He adds that in a large data set of thousands of papers even a small margin of error would leave out many preprints from the analysis and possibly skew the result. Rust says it is important that scientists read papers and evaluate the work themselves, rather than judging the quality of the paper on the basis of whether it was published as a preprint or not. He also suggests using extra caution when using assessing preprints to help make decisions in clinical settings.

Anthropic released Claude Science in June. It joins a host of other artificial-intelligence tools for researchers.Credit: Blossom Stock Studio/Shutterstock In 2010, Euan Ashley, a geneticist and cardiologist at Stanford University in California, led the first clinical analysis of a human genome, which took his team of 31 scientists nine months to complete1. This week, while unpacking after a holiday, Ashley asked the AI tool Claude, developed by Anthropic in San Francisco, California, to examine his own genome to the same standard. The analysis took 30 minutes and correctly identified an Alzheimer’s disease risk allele and gene variants affecting drug metabolism (Ashley had analysed his genome in 2012 but did not publish the results). “There is no world in which this is not utterly remarkable,” Ashley wrote in a LinkedIn post. On 30 June, Anthropic unveiled a platform called Claude Science, designed with biology research firmly in mind. The tool joins a department’s worth of general purpose AI tools for science created by technology firms and academic laboratories. Others include offerings from OpenAI in San Francisco and Co-Scientist from Google DeepMind in Mountain View, California. Another is an open-source tool called Biomni, developed by academic researchers and described yesterday in Science2. And there are many others, researchers say. “Work that usually takes me hours now takes minutes. I can really spend my time on the science that needs a human,” says co-author Yuanhao Qu, the co-founder and president of Phylo, a start-up firm in South San Francisco, California. What are these and how are scientists using them? Sometimes called 'AI scientists', these tools are based on the large language models that power chatbots, helping scientists with tasks such as literature reviews, data analysis, figure generation and manuscript preparation. They are a form of agentic AI, in which requests are broken down into steps that often involve recruiting external software systems. These scientific agents are distinct from more specialized research tools, such as the AlphaFold protein-structure-prediction model, but they can employ bespoke models. For example, Gabriele Corso, co-founder and chief executive of the London-based firm Boltz, and his team tasked a Claude agent to design an antibody that recognized two therapeutic targets, using the company’s open-source AI tools for protein-folding prediction and design. AI ‘scientists’ joined these research teams: here’s what happened The AI’s outputs aligned with the protein designers’ intuitions; Corso's team did not validate the designs experimentally, but other antibodies made with AI agents have been, he says. Boltz’s tools are among the dozens of specialized software systems that Claude Science and other AI scientists can interact with. Clare Bryant, an immunologist at the University of Cambridge, UK, was an early adopter of Co-Scientist, which mines the scientific literature and other sources to come up with scientific hypotheses. Bryant, who was investigating immune responses to zoonotic pathogens, provided the tool with a grant application and further data. Some of the ideas that it generated weren’t doable, but others were right in her lab’s wheelhouse. Her team is now testing an idea from Co-Scientist, introducing specific mutations into an innate-immune protein and seeing how they impact influenza infection. Bryant says she might have eventually come up with the experiment on her own, but it could have taken two years. “You feel like you’re talking to an oracle,” says Gary Peltz, a biomedical scientist at Stanford, who used Co-Scientist to identify existing drugs that could treat an organoid model of a disease called liver fibrosis3. How should scientists decide which tools to use? Many scientists already use AI tools such as Claude to generate presentation slides and draft e-mails. But Ashu Singhal, president and co-founder of the cloud platform Benchling in San Francisco, estimates that less than 20% of labs have fully embedded AI scientists into their research. “It’s really important that people actually try these things out, rather than simply trusting what gets shared in headlines,” he says. Singhal recommends that researchers trial several tools to work out which ones are suitable for which tasks. Hypothesis-generating AIs, such as Co-Scientist, might help during the earliest stages of a project. Later on, tools such as Claude Science and Biomni could carry out specific tasks, such as genomic data analysis. Corso recommends that researchers start with small tasks, the output of which can be verified easily. “Worst case, you have to do them over,” he says. How can researchers trust 'AI scientists'? This week’s Science paper showed that Biomni performed as well as experts on certain tasks, including rare-disease diagnosis and and cell-sequencing data analysis, often in a fraction of the time. But it didn’t do as well at some other tasks that required deep biological thinking, the researchers say. Anthony Gitter, a computational biologist at the Morgridge Institute for Technology in Madison, Wisconsin, says he has been impressed by Biomni. “Some things I wanted to try with it worked well, and some crashed pretty badly,” he says. How to build an AI scientist: first peer-reviewed paper spills the secrets To further test the tool and several other general-purpose AI tools, he entered them in a 2025 community challenge aimed at predicting important properties of preclinical drugs, such as toxicity. Entries were judged by how closely predictions matched unreleased experimental data provided by a biotechnology company. The AI tools did okay, but not as well as human experts, including those at big pharmaceutical companies. “They were doing generally reasonable things,” he says. “But it was not pushing the boundaries.” These science AIs don’t always need to be right for them to be useful. But one of their most important features, researchers say, is explaining how they arrived at a particular output, be it a hypothesis or the results of an analysis. “When using any of these tools, it’s my responsibility to know what claims are being made and how those claims are being supported,” Gitter adds. In his experiments, some models are better at this than others. Are there any downsides to using an AI scientist? For Ashley, agentic AI tools have been revolutionary to his research. “The productivity increase is enormous, like unspeakably huge,” he says. But he and other scientists Nature spoke to have deep worries about the negative consequences of AI scientists, particularly for early-career researchers and even experienced scientists working outside their fields. “You only learn to see certain types of mistakes by having made them yourself many times in the past,” says Gitter. Bryant encourages her trainees to use tools such as Co-Scientist to help form research ideas, but she hopes that she can teach them when to push back on AI use. “I have the experience to know when something’s not true.” says Bryant. “The biggest question I have is, how do I train students.”

Download the Nature Briefing Podcast 10 July 2026 In this episode: 00:29 Dolly the sheep’s 30-year legacy Metro: Dolly the sheep at 30: The clone that changed science (and celebrity petdom) Nature: From cloning to gene-editing: the enduring legacy of Dolly the sheep 05:20 The ocean floor caught in the act of splitting at the seams Nature: Ocean floor witnessed splitting apart for the first time — releasing lava Subscribe to Nature Briefing, an unmissable daily round-up of science news, opinion and analysis free in your inbox every weekday. Never miss an episode. Subscribe to the Nature Podcast on Apple Podcasts, Spotify, YouTube Music or your favourite podcast app. An RSS feed for the Nature Podcast is available too.

The US National Science Foundation, one of the world’s biggest funders of fundamental science, is expected to slash the spending of most of its directorates.Credit: IB Photography/Alamy The US National Science Foundation (NSF) is planning to expropriate money from its core science programmes to fund an initiative from the White House Office of Science and Technology Policy (OSTP), Nature has learnt. The move would strain budgets that are already tight and force the agency to rescind funding for research proposals that are nearly finalized. NSF staff members — who asked to remain anonymous out of fear of retaliation — and an internal NSF ledger seen by Nature suggest that the NSF plans to claw back around US$500 million that has already been distributed to grant-making divisions. That money would be withdrawn from three of the NSF’s eight main areas, or directorates: engineering; computer and information science and engineering; and mathematical and physical science. How Trump is following Project 2025’s radical roadmap to defund science To accommodate the expected funding withdrawal, programme officers, who handle grant proposals, at the three directorates are being required to pull back proposals that have already passed peer review and have been recommended for funding. More than 100 proposals have been affected; many of the researchers who submitted those proposals had already been informally notified of their awards, according to agency staff members. The revocation of funds would add to the squeeze on the NSF. The US Congress trimmed the agency’s total funding for this year by roughly 3% from 2025 levels, to $8.75 billion, but internal budget numbers seen by Nature indicate that about $1 billion of that money never made it to the directorates. News of the withdrawal of funds from NSF directorates was first reported by Science, which stated that the funds were going to the NSF X-Labs initiative, a ten-year programme to support technology development. An NSF spokesperson did not comment on the amount of withheld or withdrawn funds and did not specify how any such funds would be spent. But the spokesperson said that reports of funds being redirected to X-Labs are “simply wrong” and noted that funding does not expire until the end of fiscal year 2027. “Proposals that are received but not awarded remain eligible for future consideration, including in Fiscal Year 2027, unless or until they are declined or returned,” the spokesperson said. Trump’s AI ‘Genesis Mission’: what are the risks and opportunities? Several NSF staff members told Nature that at least some of the funds will be funnelled to another project, a brainchild of the White House OSTP. The OSTP did not respond to Nature’s request for comment. The White House Office of Management and Budget (OMB) has singled out the NSF for “wasteful spending” and a “growing failure of objectivity”. Dan Reed, a computer scientist at the University of Utah in Salt Lake City says that he’s concerned that money will be redirected from peer-reviewed funding channels in the NSF by the White House. Reed was the chair of the National Science Board from 2022 to 2024, which oversaw the NSF until the White House dissolved the board in April. Looming deadlines In February, the OMB sent the NSF billions of dollars that the US Congress had appropriated. But when the NSF distributed funding to its directorates in mid-April, it withheld nearly $1 billion of that money, according to an analysis by Nature of internal budget documents. At the time, staff members were not concerned because the agency frequently withholds 10 to 15% of its budget until both the OMB and Congress sign off on a detailed spending plan for the NSF. They usually do so soon after Congress sets the agency’s budget for the year. With the 2026 fiscal year ending on 30 September, NSF employees are now concerned that the $1 billion of withheld funds will not be released. If they are not, the combination of withheld funds and clawbacks would result in funding cuts of more than 30% for most directorates compared with their planned spending for 2025, according to Nature’s calculations. The exception is the technology, innovation and partnerships directorate, which is still set to receive a boost of around 30% (see ‘Funding shortfalls’). Source: NSF; Nature analysis These cuts could pit the White House’s science and budget offices against Congress. In a non-binding document that accompanied the spending bill for 2026, Congress directed the NSF to “equitably distribute funding” and avoid cutting any directorate by more than 5%. It’s unclear whether Congress has signed off on a spending plan for the NSF. Staff members of the US House of Representatives and Senate appropriations committees, which oversee such spending plans, did not respond to several requests for comment. Effect on grants To get grants out of the door before the end of the fiscal year, the NSF aims to send recommended proposals to its Office of Award Management by early August. As that deadline approaches, directorates have spent an average of 27% of their historic totals for this time of year, according to public data compiled by Grant Witness, a non-profit project that tracks changes to research funding (see ‘Far behind’). Source: Grant Witness Spending across the agency has been slowed by a variety of factors, including a 43-day government shutdown last year, a delay in distributing funding to directorates until April and budget uncertainties. Even though programme officers have less to spend than they had expected, those who spoke to Nature are unsure how many new grants will make it out the door. “They could make it through if the process is allowed to work without interference or interruption,” one staff member says. But most of “the process is now a black box and unpredictable”. Unspent money should be available until October 2027, but staff members are worried that it could be allocated elsewhere or sent back to the OMB. NSF staff members estimate that if the withdrawals are finalized, hundreds more proposals that have been recommended for funding across the engineering, computer and information science and engineering, and maths and physical science directorates would need to be sent back to programme officers for revision. Some proposals would be held until a later date when funds are available. Others would have their budgets reduced, and some would simply be declined. Staff members say that they are frustrated by the planned diversion of funds and the lack of communication about the agency’s spending. “We don’t know where the money’s going or what’s going on,” says one staff member. Programme officers are not allowed to pass on what they know to researchers. “We cannot communicate to the community at all. We’re forbidden.”

Chemist Omar Yaghi has accepted a full-time position as a researcher at Tsinghua University in Beijing, China.Credit: Justin Sullivan/Getty Nobel-prize-winning chemist Omar Yaghi has left the United States for a full-time position at Tsinghua University in Beijing, China, where he will lead a new artificial-intelligence-assisted materials discovery institute. Chemistry Nobel for scientists who developed massively porous ‘super sponge’ materials The move, first reported by the South China Morning Post, comes as the administration of President Donald Trump continues its attempts to slash US science spending and limits international research partnerships. Some nations, including China, have responded by trying to lure US talent with the promise of money and support. Earlier this year, for instance, France announced that it would award funds to dozens of US scientists relocating there. China has been wooing international researchers with talent-recruitment programmes, and some of its cities and provinces are even offering researchers lump sums and monthly allowances to relocate within their borders. Yaghi already had a connection to Tsinghua University — he became an honorary professor there in 2022. But he was officially welcomed as a full-time faculty member at a 3 July ceremony. Yaghi declined to comment to Nature for this story. However, in a recent interview with Scientific American he said that the current state of US science is “not so encouraging because of the cutting back on grants” and because of a drop in the support from US science agencies that academic researchers rely on. He also worried that US researchers were not embracing what he sees as an “AI revolution”. Researchers need to engage with AI models, he said, “as a matter of survival of the advanced research system in the US”. A materials pioneer Born in Amman, Jordan, to Palestinian refugees, Yaghi came to the United States at age 15 and had lived there until the recent move to China. He is best known for developing metal-organic framework (MOF) compounds, which are highly porous materials that have vast internal surface areas making them capable of storing gases, serving as catalysts for chemical reactions and more. Chemists have created more than 100,000 types of MOF, with an eye towards putting them to use in broad commercial applications, including harvesting water from the air and delivering drugs inside the body. Metal-organic framework (MOF) compounds usually have metal-containing nodes (blue and red) linked by organic molecules (grey and white). The one shown here, called MOF-5, is a famous example synthesized by Yaghi’s lab. The purple sphere represents the MOF’s large central pore that can fill with, for example, gases.Credit: Thom Leach/Science Photo Library Yaghi — who had been a researcher at the University of California (UC), Berkeley, since 2012 — has earned a slew of awards for his contributions to materials science, including the Albert Einstein World Award of Science, the Wolf Prize in Chemistry and, last year, a share of the Nobel Prize in Chemistry. He has also founded and co-founded several US companies, including Atoco in Irvine, California, which is developing materials for water harvesting and carbon capture, and WaHa in Fremont, California, which has created a device that turns “humidity into pure water while cutting energy costs for climate control,” according to WaHa’s website. Yaghi had stepped down from WaHa’s board in 2022, says Frank Ramirez, co-founder and chief executive of the company, adding that its businesses will be unaffected by his move to China. As for Atoco, Yaghi’s move will keep him more involved with the company than ever before, says Samer Taha, the company’s chief executive. Atoco is collaborating with the Yaghi Science Initiative, a nonprofit launched by the Nobel laureate to connect researchers across a number of countries and to support early-career researchers in solving global challenges. Yaghi’s move to China is part of this global science initiative, Taha says, and it “will multiply the opportunities for transformative discoveries”. Tackling complex problems A possible motive behind Yaghi’s move could be that after winning his Nobel Prize, he wants to “do something more” and “build a new paradigm of research by combining AI, chemistry and material sciences”, says Marina Zhang, a science-policy researcher at the University of Technology Sydney in Australia, who focuses on innovation in China. How China is vying to attract the world’s top scientific talent Tsinghua’s goal with Yaghi at the helm of its new programme is to “tackle complex problems beyond any single field”, and to bridge “Eastern and Western intellectual traditions for the benefit of all humankind”, says Lei Liu, who is the chair of the university’s chemistry department. The future of Yaghi’s research group at UC Berkeley remains unclear. In a statement to Nature, the university wrote that it congratulated him on his move to Tsinghua and added: “Professor Emeritus Yaghi continues to hold an appointment as Professor of the Graduate School at UC Berkeley, and we look forward to his continued contributions to Berkeley’s graduate programs”. But it did not answer questions about the fate of his team. As of late June, Yaghi’s official title listed with UC Berkeley was not emeritus, but was still the James and Neeltje Tretter Professor of Chemistry. China rising China has been making a sustained effort to increase its spending on research for years, aiming to become a global leader in science. According to a report by the Organisation for Economic Co-operation and Development (OECD), China’s overall spending on research surpassed that of the United States in 2024, with the former nation spending US$1.03 trillion annually and the latter spending $1.01 trillion annually that year. World’s most porous sponges: intricate carbon-trapping powders hit the market China has been attractive to international researchers such as Yaghi because “it can provide greater latitude to pursue and accelerate AI research and framework development”, Ramirez says. He also says that current US policies around research funding and immigration are affecting the country’s ability to attract and retain top international talent. Several prominent US scientists have made a move to China in the past year. For example, Dan Yang, a neuroscientist at UC Berkeley who had spent over three decades working in the United States, joined Shenzhen Medical Academy of Research and Translation (SMART) as a senior principal investigator last year. And liver cancer researcher Feng Gensheng reportedly left the University of California San Diego in August last year to serve as director of the Cancer Research Institute at Shenzhen Bay Laboratory in China after 40 years in North America, according to the South China Morning Post. Yaghi’s move seems different, though, Zhang says. Although China had previously recruited international scientists with the primary purpose of training students, Yaghi’s appointment could establish an entirely new model of research, she says. This all shows that China’s agenda for this appointment isn’t to “catch up” with global scientific powerhouses such as the United States, “but to establish its own turf and its own technological capabilities”.