Bristol Myers Squibb will combine two AI clusters so that data from different studies can help choose the next experiment. On July 20, BMS announced the deployment of a second NVIDIA cluster. The company plans to connect it to the first one, so that research teams at different sites can work with common data and computing resources. A cluster is a group of connected servers on which artificial intelligence models are trained and perform calculations. BMS has been using such a system for about three years. It was difficult for scientists to access it: after the company's acquisitions at different sites, local rules remained, and special skills were required to work with it. In an NVIDIA publication on July 20, BMS described the next step: the company wants to create a single computing environment for all its research sites. The search for a drug begins with the selection of a biological target - a protein or another element of the organism that the drug should act on. Then, scientists select molecules, test them in the laboratory, and decide which experiment to conduct next. BMS reports that it is already applying the Predict First approach: a model prediction helps plan a laboratory experiment before scientists start it. When BMS combines the systems, the data from the research program in Lawrenceville, New Jersey, will be able to become the input for models used by the team in San Diego. The company also plans to give scientists the opportunity to run complex predictions in plain language. The result of one study should become the material for the next hypothesis in another. The second cluster is still being deployed, and the integration of the two systems remains a BMS plan. The company has not published comparisons that can be used to judge how the common environment affects the choice of experiments, program timelines, or the number of candidates that reach clinical trials.

🔗 Read original →

Levent Alpöge published a counterexample to the Jacobian conjecture in three dimensions and thanked Fable for work on the problem. On July 20, mathematician Levent Alpöge published a formula for a map of three complex variables. He thanked Akhila Mathew for the question and Fable for the work. The Jacobian determinant of this map is always -2, but three different source points yield the same result. The Jacobian conjecture, formulated by Keller in 1939, concerned polynomial maps: sets of formulas that assign another triple to each triple of complex numbers. For such a map, a table can be compiled showing how each result changes with a small change in each input; the number calculated from this table is called the Jacobian. The conjecture claimed: if the Jacobian is constantly non-zero, the map should have an inverse polynomial formula. The value of -2 means that near each source point, a nearby result can be uniquely restored to a nearby source point. One inverse formula for the entire space requires more: each result point should correspond to exactly one source point. In Alpöge's post, he cited three different source points: (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2). The map sends each of them to (-1/4, 0, 0). One result has three different preimages, so the map does not have a common inverse formula. The mathematical note reduces the search for preimages of an arbitrary result point to a cubic equation, i.e., an equation of the third degree. For (-1/4, 0, 0), it factorizes to -s(s-2)(s+2)/2. Its roots 0, 2, and -2 exactly correspond to the three published points. This map has sequences of source points that go to infinity, although their results converge to a finite point. The constant Jacobian controls the map near each point, but not throughout the space. In the original post, Alpöge described the contribution of the participants as follows: "The Jacobian conjecture is false. Thanks to my close friend Akhila for the question and my other close friend Fable for the work during the World Cup final." The published materials allow independent verification of the formula and its properties. The publications do not reveal who made each step in the search. A week before that, physicist Yuji Tachikawa reported that Claude Fable had found a computational error in a stuck problem in string theory and developed its course. In the current case, the publication provides a formula, a constant Jacobian, and three points that can be verified separately. If new coordinates are added to this map and left unchanged, the counterexample works in any number of variables starting from three. The two-dimensional case of the Jacobian conjecture remains open.

🔗 Read original →

Deep Origin has retrospectively ranked TCIP3 as the top molecule among 17 molecular glues - compounds that bring two proteins together. TCIP3 connects BCL6 with p300/CBP - enzymes that help the cell turn on genes. In lymphoma cells, this binding triggers programs that stop growth and kill the cell. Deep Origin checked if its calculation matched the molecule that the laboratory had already identified as a leader. On July 20, an article by researchers from Stanford, MD Anderson, and Deep Origin about TCIP3 was published in Cell. In some cases of diffuse large B-cell lymphoma, BCL6 keeps genes turned off that slow down cell division and trigger cell death. TCIP3 binds to BCL6 with one part and to p300/CBP with another, bringing these enzymes to the DNA regions controlled by BCL6. TCIP3 brings p300/CBP to the genes that BCL6 keeps turned off. The enzymes acetylate BCL6 and the proteins that DNA is wound around in these regions. After this, BCL6 is less able to keep genes turned off, and the cell turns on programs that stop growth and apoptosis - controlled cell death. An open preprint by the authors in 2025 showed this mechanism in cell lines. In a 72-hour test on lymphoma cells, TCIP3 suppressed their growth with an IC50 of around 0.8 nanomoles per liter: half of the maximum suppression was achieved at a concentration of less than one billionth of a mole. Control compounds with the same chemical linker that could not bind to either BCL6 or p300/CBP were more than a thousand times weaker in toxicity to these cells. Thus, the authors showed that the effect requires the triple assembly of BCL6-TCIP3-p300/CBP. In the July 20 announcement, Deep Origin described its calculation. The company built a triple complex for each of the 17 compounds, simulated the movement of these assemblies, and calculated the stress of the chemical linker between the two parts of the molecule. This linker affects whether the proteins can maintain the necessary mutual arrangement. In Deep Origin's ranking, TCIP3 was first in a retrospective calculation: its laboratory activity had already been measured before this analysis. In the international CASP competition, teams submit predictions of protein structures that experimenters have already measured but have not yet disclosed. A similar test for Deep Origin would look like this: the calculation ranks previously unmeasured compounds, the laboratory synthesizes several candidates from the top of the list, and then measures their activity. Only Deep Origin reports on the almost complete disappearance of tumors in mice with xenografts - transplanted human tumors. The open preprint from 2025 calls TCIP3 a tool molecule and says that the authors did not study its efficacy in a live mouse model of this lymphoma at that time; the full text of the Cell version from July 20 is not available. Even successful ranking of a new series will show a prediction for a chosen laboratory system, and the connection between cellular and animal models and the result in patients will require other data.

🔗 Read original →

Retro obtained a transplant from six adult donors' cells, which twice restored human hematopoiesis in mice and preserved a DNA methylation age of around five years. On July 17, the Retro team posted an unrevised preprint on donor cells aged 18–60 years. After the second transplant, 18 of 23 immunodeficient mice had more than 1% human cells in their bone marrow; Horvath clocks estimated the age of isolated human hematopoietic cells with the CD45 marker at 5.28 ± 1.45 years. Blood stem cells constantly replenish erythrocytes, platelets, and immune cells. After transplantation, they should settle in the bone marrow and produce all these lines for a long time. With age, this becomes more difficult. Reprogramming an adult cell into an induced pluripotent stem cell, or iPSC, resets many DNA methylation age marks and lengthens telomeres. For transplantation, such a cell needs to be converted into a blood stem cell and tested in a living organism. Resetting age marks does not eliminate all damage: in serially cloned mice, by the 58th generation, cloning success sharply decreased due to accumulated genomic errors. In 2024, the Elizabeth Ng group obtained iPSC-derived cells that engrafted in the bone marrow of mice for a long time and produced several blood lines. In April, Retro described the path from iPSC to hematopoietic stem cell transplantation. The new preprint checks two things: whether such cells can re-engraft in the bone marrow and maintain a young DNA methylation age pattern. The team obtained iPSC-derived hematopoietic stem cells, iHSC, in 15 days and introduced them into immunodeficient NBSGW mice. In such animals, human transplants can engraft in the bone marrow. After 20 weeks, the bone marrow of recipient mice from three donor lines was transplanted into other mice. In the second cycle, 18 of 23 animals had more than 1% human cells in their bone marrow; these cells again formed erythrocytes and several types of immune cells. In the first transplant, each mouse received five million iHSC, while control animals received from 50,000 cord blood cells to one million adult hematopoietic stem cells. This design shows that cultured cells can restore blood at this dose but does not compare the potency of one iHSC to a conventional transplant. During 15 days of iHSC differentiation, the pluripotency program was turned off and the hematopoietic program was turned on, but their methylation pattern still differed from that of adult blood stem cells. After living in the mouse bone marrow, it became closer to the adult profile. Methylation was studied in a mixture of isolated human cells with the CD45 marker. Therefore, the adult profile may indicate both cell maturation in the bone marrow and the growth of a subset of cells that initially engrafted better. The methylation pattern answered two different questions: did the cells become similar to adult blood stem cells and did they preserve a young age according to DNA methylation marks? Horvath clocks translate the methylation pattern into an age estimate. After the second transplant, they gave 5.28 ± 1.45 years. In this mouse model, cells from adult donors acquired signs of adult hematopoietic transplants and preserved a young age according to DNA methylation marks.

🔗 Read original →

The White House is proposing to test the rules for awarding scientific grants with an experiment. On July 21, the White House Office of Science and Technology Policy released a report, Science: A New Golden Age, and a memorandum on research and development budget priorities for fiscal year 2028. Agency heads with research and development budget authority of at least $3 billion in fiscal year 2026 must submit an action plan to the Office of Science and Technology Policy and the White House Office of Management and Budget within 90 days. The agencies should consider these priorities in their fiscal year 2028 budget requests. The US federal government spends around $200 billion on research and development each year. The report proposes to test the rules by which the state selects recipients of scientific funding. To do this, the memorandum suggests that agencies develop metascience - research on how agencies select applications, review them, award grants, and obtain results. It proposes linking data on applications, reviewer evaluations, grant decisions, and subsequent outcomes in one system. This would allow for testing a different selection method on parts of the federal research portfolio and comparing it to the usual method based on subsequent outcomes. The proposals include portable fellowships that remain with the researcher when moving to another organization, and rapid grants for preliminary and exploratory work. A rapid grant requires only a few pages of application, and a decision should be made in less than a month. A "golden ticket" would allow a technical reviewer of an agency to recommend an unusual application that an expert panel did not support. New mechanisms, including the "golden ticket", the memorandum proposes to study, pilot, and evaluate. AI for science can quickly propose a molecular target, material, or explanation of a biological process. A laboratory then conducts an experiment and checks the prediction. The journal Science has already warned that AI can produce more science than people can check. The report proposes combining models with open data, robotized laboratories, and reproducibility packages - sets of data, code, and analysis conditions with which another laboratory or program can reproduce the result. The Genesis Mission is a federal AI for science program that combines data, computing, and research equipment. In March, Argonne presented more than a dozen Genesis projects, including AI for enzyme research and a network of autonomous laboratories. On July 22, the White House announced more than $5 billion in federal commitments and 278 selected projects in the Genesis Mission. Among the tasks of Genesis is to "help Americans live longer and healthier". To do this, the Department of Health, the Department of Energy, and the Department of War of the USA plan to combine data on molecules, genes, organism signs, treatment, and real clinical practice to search for new applications of existing drugs and to bring new therapies to patients faster.

🔗 Read original →

In four patients with complete cervical spinal cord injuries, no tumors were found after transplantation of neural cell precursors over 2-4 years. The results of the first trial of such transplantation in humans were published in Nature Medicine. Four men were administered two million cells each, 14-28 days after injury; the goal of the first phase is to test the safety of the procedure. In complete cervical spinal cord injuries, signals from the brain do not pass to the body below the injury. In double neural bypass, implants read the desire to move an arm and, with electrical stimulation of muscles and spinal cord, bypass this gap. Here, surgeons introduced cells directly into the damaged spinal cord segment. The transplant consisted of neural precursors - cells from which neurons and cells covering nerve fibers with a protective sheath - myelin, can develop. They were grown from iPSC - reprogrammed donor umbilical cord blood cells, which can again transform into different tissues. One cell line allows for the pre-manufacture of identical batches for several patients. Before transplantation, the laboratory checked the cells for genetic changes and pushed them to a more mature state to reduce the risk of excessive growth. Then, patients took tacrolimus for nine months - a drug that suppresses the immune system and prevents the body from rejecting the donor cells. Over the first year, doctors recorded 84 undesirable events: two moderate surgical complications and 22 mild or moderate reactions related to tacrolimus. MRI and positron emission tomography (PET), which shows tissue metabolism, did not reveal signs of tumor growth at the transplantation site. Over 2-4 years of observation, none of the four patients experienced severe reactions related to the cell product, and neurological functions did not worsen. In all participants, scores on movement and self-care scales increased over the year. Two patients developed voluntary movements below the injury. These improvements cannot be attributed to the transplantation without comparison with similar patients who did not receive cells. In the first months after injury, some functions return on their own. The next study should distinguish between natural recovery and the effect of cells. The current work answered a different question: can a uniform cell product be prepared, introduced into the damaged spinal cord, and patients be observed for several years without signs of tumor growth.

🔗 Read original →

On July 17, Irina Conboy's group published a study on DMA, a combination of dichloroacetate, metformin, and navitoclax. The treatment began at approximately 18 months of age. In the nine mice receiving DMA, the median lifespan was 1002 days, while in the nine control mice, it was 815 days. After damage or other stress, some cells stop dividing, remain in tissue, and release inflammatory signals; these cells are called senescent. Navitoclax blocks BCL-XL, a protein that helps both senescent cells and platelets survive. Platelets help stop bleeding by forming a clot, and the same protein is necessary for the survival of both senescent cells and platelets. 18 hours after administering 50 mg of navitoclax per kilogram of body weight, the number of platelets was approximately one-quarter of the control. The authors reduced the navitoclax dose to 5 mg per kilogram and added metformin and dichloroacetate. Metformin inhibits complex I in mitochondria, one of the links in energy production. Dichloroacetate helps direct pyruvate, a product of glucose breakdown, into mitochondria. The authors tested the hypothesis that senescent and cancer cells are worse at switching between ways of obtaining energy, so the two drugs would create a burden for them and increase their sensitivity to a low dose of navitoclax. In cultures, DMA almost reduced ATP, the molecule that transports energy within cells, to background levels in senescent cells of connective tissue and MCF-7, a line of breast cancer cells. Healthy connective tissue cells maintained their ATP level. Exogenously added ATP partially restored the viability of senescent cells. Partial restoration of viability supports the connection between the decline in ATP and cell death. In human neural precursor cells, viability did not decrease. In human muscle cells, it decreased by approximately 20%, and in liver cells, by 5%. From these cultures, it is impossible to understand which cells DMA affects in mouse organs. The median lifespan in the DMA group was 187 days longer, and the average total lifespan was 12% longer. A separate analysis of males and females did not allow confidently distinguishing the effect of DMA from random fluctuations: there were three to six animals in each subgroup. In the survival experiment, the authors tested the effect of the entire DMA mixture. Control mice received a solution without drugs; there were no separate groups for each component in the experiment. The mice were observed until natural death, unlike the Immorta Bio study, where survival was measured after a toxic regimen of doxorubicin in young mice. In this group, the authors did not measure the number of senescent cells in organs. One protein marker does not provide such a map: the SenNet atlas showed that the set of senescence markers changes with cell type and cause of aging. The survival data relates to the DMA mixture as a whole. After DMA, the number of platelets was approximately 70% of the control; the authors were unable to distinguish the difference from random fluctuations. This short test only measured the number of platelets. In several animal models of pulmonary hypertension, elevated pressure in the lung vessels, ABT-263 was accompanied by the loss of cells lining the vessels and bearing signs of senescence, as well as worsening of lung vessel function. In the DMA study, the authors did not measure which cells the mixture affects in mouse organs and how this affects their function.

🔗 Read original →

Alcor announced that it began training local European deployment teams for cryopreservation on July 21. Alcor reported that preparation has been declared for brain and whole-body cryopreservation. The organization does not disclose the countries, the number of trainees, the date of the first independent deployments, and the measured response time. After the cessation of blood circulation, the brain stops receiving blood, oxygen, and glucose; neuronal damage increases within minutes. DART is Alcor's deployment team: when a member's life is expected to end, it can be on standby at the bedside, and in the event of sudden death, it must arrive at the patient as soon as possible. After confirming legal death, the team performs primary stabilization and organizes transportation to a storage facility. In its July bulletin, Alcor reported that its DART team is already deploying to organization members worldwide and that it has begun training local teams in Europe. For sudden cases, a team from the US has to cross the Atlantic. Alcor expects that locally trained specialists will reduce this delay. In April, Alcor gathered more than thirty specialists from the US, Canada, and Europe in Arizona for four days of training, practice, and drills. According to Alcor's internal rules, an active DART member undergoes annual recertification: confirms a minimum number of actual deployments and passes an exam. European teams, according to Alcor's statement, will be trained to the same standards and goals. In its May bulletin, Alcor reported on James Arrowood's trip to Europe and meetings with potential partners. In July, the organization announced that it was already training local teams. On the Alcor Europe page, the legal structure, headquarters, and long-term storage are described as future projects after securing funding. Training local DART teams relates to the earliest part of the procedure - the time from legal death to primary stabilization.

🔗 Read original →

A team led by David Liu, with the help of AI, strengthened proteases so that laboratory evolution could find new functions. On July 22, a study by David Liu's team was published in Nature. ProteinMPNN redesigned three proteases, and automated laboratory evolution taught them to cut new protein targets. The best variant against ataxin-2 turned out to be more than 79 times more selective than the variant grown from a natural enzyme. To teach a ferment to cut a new target, mutations must change protein recognition and preserve its three-dimensional shape. Often, a new useful mutation makes the protein less stable: it folds worse, and selection weeds it out along with the new function. Liu's team decided to strengthen the protease before selection. ProteinMPNN received three-dimensional structures of proteases of botulinum neurotoxin and modified areas away from the catalytic center. One of the proteases had 74 variants, 58 of which preserved activity, and 22 gave more soluble protein. AI created a reserve of stability, and the experiment checked which new functions this reserve allowed to survive. Then the authors launched PACE - a system of continuous evolution with bacteriophages. The phage multiplied only when the protease cut a given peptide; in a day, the system went through dozens of generations of mutations and selection. On the most difficult of the three new targets, a working ferment appeared in all four lines started from the redesigned variant D3, and in two out of four lines from the natural protease. When the authors transferred the found mutations between proteins, mutations from D3 often lost function in the natural protease. Mutations from the natural start, on the other hand, worked in D3. The same amino acid substitution behaves differently in a different protein background. This is consistent with the fact that D3 carries part of the useful but destabilizing mutations. In the final campaign, the protease was tuned to ataxin-2 - a protein associated with the risk of lateral amyotrophic sclerosis - and selection was carried out for cleavage of SNAP25, its natural target. The best variant from D3 had a ratio of ataxin-2 cleavage to SNAP25 cleavage more than 79 times higher than the best variant from the natural start. In a culture of human HEK293T cells, it gave more target product and fewer side fragments. In a May study by the same laboratory, ProteinMPNN stabilized an already evolved DNA editor PE8. In the new article, the model acts earlier: it prepares the starting ferment from which selection then obtains variants with a new function. The authors tested three related proteases, bacterial selection, and HEK293T cells. For treatment, delivery of the ferment to motoneurons, testing of the immune response, safety, and effect in animals will be required.

🔗 Read original →

Repligen is acquiring BioLife for $1.5 billion and adding cell therapy storage to its business. On July 22, Repligen and BioLife Solutions announced an agreement where BioLife shareholders will receive 64% of the deal's value in Repligen stock and 36% in cash. The companies are expecting the deal to close in the fourth quarter of 2026, following a BioLife shareholder vote and regulatory approvals. Cell therapy comes to the patient as a living material, and cells need to be processed, stored, transported, and prepared for administration in a way that ensures their survival. BioLife sells media for freezing and storing cells, as well as tools for their processing. Its CryoStor is a solution in which cell products are frozen and stored. According to the companies' statement, the CryoStor line is used in 18 approved cell therapies and in most sponsored cell product trials in the US. Repligen is acquiring a supplier of consumables for the stage at which cell products are frozen, stored, and transported. Among commercial therapies using BioLife products, BioSpace names Carvykti, Yescarta, and Breyanzi. Repligen already sells biopharmaceutical companies filters, liquid systems, chromatography equipment, and analytical instruments. After the deal closes, BioLife will add cell preservation means to this lineup; the company expects new types of therapies, including cell therapies, to account for around a quarter of its revenue. The $1.5 billion is an estimate of BioLife's entire business, not just the CryoStor medium. Repligen also forecasts $20 million in savings in the first year after the deal and $30 million in the second. These amounts are contingent on the deal closing and the companies' integration.

🔗 Read original →

MIT has directed the growth of vascular branches through magnetic stretching in a model of a human vessel. In a collagen gel, researchers changed the force and direction of stretching on the vessel wall. With weak stretching, more branches formed, while strong stretching caused individual branches to grow longer. Engineered tissue requires a network of thin vessels for cells to receive oxygen and nutrients. Large channels can be printed, but capillaries must sprout from cells and connect into a functioning network. Their shape is currently difficult to control with the same precision. On July 14, MIT released a breakdown of the research, published in PNAS on July 6. In the collagen gel, the team created a hollow channel and lined it with human endothelial cells - the cells that form the inner lining of blood vessels. Near the channel, they placed a small magnet; an external drive moved it and stretched the gel on the vessel wall for one hour a day for three days. The force of stretching changed the number and length of branches, and the direction of stretching changed their route. With stretching at 5% of the gel's width, more branches formed. At 15%, there were fewer, but individual branches were longer. When the researchers changed the direction of the magnet's movement along three axes, the branches turned to follow the stretching; some became L-shaped. Within some branches, a channel connected to the original vessel was preserved, which the authors verified using a fluorescent dye. This control currently only works on the early growth of branches in a model of a single vessel. The authors have not yet assembled a complete vascular network. The next test is to obtain a dense network from these branches that can supply complex tissue for a long time and connect to the bloodstream after transplantation. The researchers also tested the cells' response to stretching. When they suppressed PIEZO1 - a gene that encodes a stretch-sensitive ion channel - fewer branches formed. However, stretching still maintained the barrier function of the vessel wall, as other mechanisms also participate in the cells' response to force.

🔗 Read original →

US federal agencies searched for active scientific grants by keywords before terminating funding. Agreements with federal agencies, attached to a court petition, describe the general procedure: first, grants were found by words and themes, then some of them were selected for termination of funding. In individual cases, artificial intelligence tools were used when compiling lists. On July 15, the plaintiffs in the case of Thakur v. Trump filed a petition in federal court and attached agreements with agencies. The National Science Foundation, the National Foundation for the Humanities, the Department of Defense and Transportation, and the US National Institutes of Health described in them how they selected grants for consideration for termination. "Search terms, keywords or phrases" highlighted grants for review. According to the text of the agreements, then projects that expressed or allegedly expressed positions not supported by the administration were selected from the list. The agencies applied general criteria and standard letters, rather than separately checking whether each grant recipient had met the conditions. The same agreements state that the recipients did not violate the terms of funding. At the National Institutes of Health and the Department of Health, the list of themes was expanded: from projects on diversity, equality, and inclusion to gender, vaccine skepticism, and COVID-19. The lists included "health equity", "structural racism", and "sexual orientation". The materials do not explain what data the AI tools processed and how their results influenced the decisions. The agencies only indicated that AI could participate in preparing the lists provided to it. The Associated Press reported on July 21 that the case concerns more than a thousand grants from the University of California. The plaintiffs estimate the previously awarded funding at approximately $2 billion. A hearing on the petition is scheduled for October 20, and the court has not yet decided whether this procedure is lawful. This differs from the White House plan to experimentally test the rules for issuing new grants, where it is proposed to compare the method of selecting applications with research results. The Thakur agreements describe the review of already issued grants by thematic words and the alleged position of the project. For long-term biomedicine research, continuity is important: a team is hired for a grant, participants are recruited, and experiments are conducted for years. The materials do not allow us to establish the consequences for each laboratory, but the described procedure creates a risk for such programs: the funding of an already started study may be revised without a separate assessment of whether the laboratory has met the conditions of the grant. This applies to aging research as one of the long-term biomedical areas; individual gerontology programs are not named in the documents.

🔗 Read original →

On July 22, Haowei Man's team published an article in Nature about ContactSeek. The method combines AlphaFold3 predictions for the protein-RNA-DNA complex with sequencing data and suggests amino acid substitutions for more selective DNA editors. A base editor changes one DNA letter without a double-strand break. A guide RNA leads it to the desired sequence, a CRISPR protein holds the complex on the DNA, and an attached enzyme performs the chemical substitution. Similar sequences sometimes also hold this complex, and then the editor changes letters outside the chosen target. Usually, engineers screen amino acid substitutions in the protein and measure the result. ContactSeek starts with the error traces of the original editor: the authors found genome-wide sites where it came together with the guide RNA and submitted the target and these site sequences to AlphaFold3. The model built complex variants from protein, RNA, and DNA. Here, AlphaFold3 builds a map of likely contacts within the complex. ContactSeek compares the differences between the target and off-target sites with the sequencing signal. This is how the program identifies amino acids whose contacts with RNA or DNA change along with the frequency of off-target edits and suggests substitutions for cellular experiments. The authors tested this route on the adenine editor Cas9 and the cytosine editor Cas12a in HEK293T cells. The ABE8e-DD variant had a total guide-dependent off-target editing signal 99.2% lower than the original ABE8e: on the ABEsite16 target, the comparison covered 270 off-target sites. For the cytosine editor variant on Cas12a, the signal decreased by 82.1-95.1% for four guide RNAs. The authors also measured target editing, off-target RNA changes, and editing without guide RNA. Base editors have already reached patients: in personalized therapy, the KJ Malton editor was delivered to the liver with lipid nanoparticles, and after a year, clinical improvement was noted in a child without serious side effects. ContactSeek has been tested only in HEK293T cells, so its variants still need to be tested for delivery to tissues and in long-term observations. ProteinMPNN helped obtain prime editor variants that accumulated better in cells. ContactSeek solves a different problem: it narrows down the list of places where the editor can trigger outside the target. The contact map turns a broad screening of amino acid substitutions into specific hypotheses for experimentation.

🔗 Read original →

User LessWrong bits suggested that people will start making plans for future anti-aging therapy before it appears. On July 20, bits published an essay on when radical life extension will enter people's personal plans. This refers to increasing healthy and overall life beyond what current medicine and habits provide. In the essay, bits asks when a person will decide that they will be able to take advantage of future therapy. Then the planning horizon changes: decisions about retirement, savings, children, and career depend on the new understanding of how much time is left. To explain this, bits uses the concept of Turkish-American economist Timur Kuran - preference falsification. A person may want to live significantly longer, but publicly repeat the usual norm as long as such a desire seems like a strange fantasy. When future therapy seems plausible, people start talking about it openly, and it becomes easier for the next person to do the same. This model is based on two assumptions: many people already want a long life, but hide this desire; a change in norms will allow them to speak out. In the essay, this is an explanatory hypothesis, not a measurement of public opinion: the author does not provide data on the scale of hidden demand or the speed of the cascade. In his scenario, the cascade can be triggered by signals that are easy to see and retell: a recognized scientific result of rejuvenation in humans, a notable discovery using AI, a new therapy in the public eye, or a regulator's permission to measure aging in clinical trials. Such signals make future therapy a subject of ordinary conversation. Then, bits suggests, demand for research, money, and political decisions change. In a comment to the essay, Dagon suggests a different sequence. "The strongest factor in expected lifespan - measured or at least claimed life extension or significant health extension in old age," he writes. A wide audience, in his opinion, will change expectations after a result; early supporters are able to believe earlier and be wrong about the timing. The dispute concerns the order of events. According to bits' model, public expectation is able to gather support for future research. According to Dagon's model, a visible result appears first, which people trust. If people start changing plans before such a result, bits' hypothesis will gain support. If expectations shift after results, Dagon will be proven right.

🔗 Read original →

Colossal is in talks for funding at a valuation of $20-30 billion. On July 20, Axios reported that Colossal Biosciences is in negotiations for new funding at a valuation of $20-30 billion. In January 2025, the company raised $200 million at a valuation of $10.2 billion; Axios writes that over the past year, it has begun to generate revenue. Colossal is known for its programs to bring back the mammoth, dodo, and woolly wolf. At the same time, the company is building biobanks, genetic medicine programs, and reproductive technologies. These areas have grown out of work with the genomes of extinct animals, but they already have their own customers and products. Axios writes that Colossal is creating repositories of genetic material of endangered species together with the US government and the Dubai Museum of the Future. In June, a memorandum with the U.S. Fish and Wildlife Service on a cryobank of tissues, cells, and genomes gave this work a specific shape. This is a conservation service that can be developed in parallel with long-term de-extinction programs. Colossal also has a biomedicine line. Form Bio, spun out of it, helps teams creating genetic medicines: selecting designs for gene therapy and parsing production data. TechCrunch also mentions companies Breaking and Astromech, which have grown out of Colossal. Thus, work with animal DNA is being transformed into tools for drug development and services for government conservation programs. According to the author's model, the $20-30 billion range refers to a company where de-extinction remains the most notable project, and biobanks, genetic medicine services, and spun-out companies are already working alongside it. The negotiations assess the entire set of technologies, not just the promise of one day bringing back an extinct animal.

🔗 Read original →

Antonio and Pablo Acuaviva described five results in Banach space theory for which the model proposed ideas and drafted proofs. On July 19, Antonio and Pablo Acuaviva posted a preprint on five results in Banach space theory. The authors write that the model found key moves and wrote draft proofs, which they then checked, corrected, and formalized into final arguments. Banach spaces allow for measuring distances between objects and working with infinite sequences; in this language, mathematicians describe functions and function transformations. One missing hypothesis here breaks the proof, so the authors read each step, corrected errors, and verified the arguments against previous theorems. In four of the problems, according to the Acuavivas, the model's original answers already contained the basic proof: the authors had to correct references, individual errors, and the presentation. In the fifth, the model proposed a plan for a long proof, and the authors filled in the transitions and assembled a coherent argument. In the same work, the authors describe the search for problems in the scientific literature. Scripts took original article texts from arXiv and marked phrases like "question", "problem", and "hypothesis". Then, an agent read a fragment, searched for a proof, counterexample, or already known answer, and saved the result in a package for a mathematician. The package contains the original article, the exact formulation of the question, the course of the argument, and the found references. The main run went through a queue of 1,433 articles on functional analysis. Mathematicians spent individual packages on parsing: 31 received the status of verified, 10 were rejected. The five results from the first part of the preprint grew out of problems that the authors chose themselves, not from this queue. Two different modes emerged: in one, the model works on a task selected by people, and in the other, it searches for questions in articles and prepares material for analysis. In the story with the lower bound for convex optimization, the argument is recorded in a formal language, and Lean checks it line by line. With the Acuavivas, the proofs are checked by the mathematicians themselves: their package assembles material for analysis, rather than replacing verification with a formal certificate. With FunSearch from Google DeepMind, the model also proposes options, but in the form of programs: an automatic evaluator launches the code and selects successful ones. A proof cannot be launched in this way: a mathematician must check each step, match the theorem with the question and with the published literature. The model quickly iterates through moves, and the mathematician chooses a question, checks the argument, and searches for its place among known results. In this work, the model's answer turns into a research result only after such verification.

🔗 Read original →

Kolon TissueGene reported preliminary results of the ACTiVION-II trial on July 20: in 531 patients, a single injection of TG-C did not improve knee pain and function better than saline placebo after one year. The company also reported that the therapy did not achieve all pre-specified secondary endpoints. TG-C combines donor cartilage cells and genetically modified cells that produce TGF-β1 protein. Developers tested whether a single injection into the joint could alleviate symptoms of knee osteoarthritis, a disease that makes ordinary movements painful and stiff. In ACTiVION-II, doctors randomly assigned patients with knee arthritis to either TG-C or saline placebo. After 12 months, they compared pain on the Visual Analog Scale (VAS) and knee function on the WOMAC questionnaire, which asks about pain, stiffness, and everyday movements. According to Kolon TissueGene's release, TG-C did not provide a statistically significant benefit in either pain or knee function. The company also reported a comparable frequency and severity of adverse events in both groups. This is a separate outcome: patient safety and benefit require different data. So far, the company has only disclosed a preliminary summary without effect size, p-values, and subgroup analysis. ACTiVION-II tested a single injection of TG-C in patients with knee arthritis. Its result describes this therapy and this patient group; data from a second Phase III trial, ACTiVION-I, Kolon TissueGene expects in October.

🔗 Read original →

Insilico Medicine released 3D-Fit for testing language models in molecule design. On July 20, the Insilico Medicine team posted the 3D-Fit preprint. In the test, language and diffusion models create molecules for 1,453 protein-small molecule complexes. The drug molecule must occupy the protein's pocket - a cavity between its amino acids. A chemist can ask to preserve a fragment of a known molecule, attach a group for a hydrogen bond at a given point, or reach a specific amino acid. Each such contact gives the entire molecule additional conditions: the bonds and angles must remain valid, and the molecule itself must fit among the protein's atoms. In 3D-Fit, the authors give the model the coordinates of the pocket's atoms and a text condition: an anchor fragment, a pharmacophore point - a place for the necessary chemical function - or a mandatory contact with the protein. The model outputs the coordinates of the ligand, i.e., the candidate molecule. Then the authors separately measure two properties. The first is whether the model fulfilled the local spatial instruction. The second is whether the molecule was successfully restored, whether its geometry withstands physical checks, and whether its pose fits in the pocket. Language models often cope with anchor fragments and pharmacophore points. The authors associate this with the task format: the text condition is similar to a record that the model should generate. But the entire pose requires simultaneously agreeing on the shape of the molecule and its position among the protein's atoms. In this test, language models gave way to specialized diffusion models. In the initial poses, all the tested language models received a UniDock score above -6 kcal/mol; the authors use this threshold for weak binding. Local optimization shifted most results to approximately -6...-7 kcal/mol. The best diffusion models received lower scores, meaning more favorable poses in this metric, and more often passed the geometry checks in PoseBusters. Local contact is not enough to make a molecule a good candidate. 3D-Fit checks whether this contact is preserved when the entire molecule passes through the constraints of chemistry and the three-dimensional protein pocket. The physical check separates hitting a given point from a molecule that entirely fits in the pocket.

🔗 Read original →

TGN-073 reduced tau pathology in mice by activating the AQP4 water channel. In PS19 mice with tau pathology, the AQP4-activating compound TGN-073 enhanced the influx of MRI tracer from cerebrospinal fluid. The brain tissue showed less pathological tau, better preservation of neurons, and weaker gliosis, a reaction of support cells to damage. This effect was not observed in mice lacking AQP4. The article was published on July 18 in Molecular Neurodegeneration. PS19 is a line of mice with human mutant tau protein that aggregates into clumps inside neurons, damaging them and reproducing part of the tauopathy picture. The authors chronically administered TGN-073 to these mice, a compound that activates the AQP4 water channel. The brain has no separate pipe for cleaning up proteins: cerebrospinal fluid moves along vessels, exchanges with intercellular fluid, and carries away dissolved substances, a route called the glymphatic pathway. AQP4 is located in the processes of astrocytes around vessels and directs the flow of water through their membranes. Therefore, the authors measured not only the total amount of AQP4 in the brain but also how densely the channel is assembled around vessels. In PS19 mice, fluid exchange weakened even before the severe stage of the disease and worsened with age. After TGN-073, more MRI tracer entered the brain from cerebrospinal fluid. The tissue showed less pathological tau, neurons were better preserved, and gliosis was weaker. At the same time, more tau appeared in cerebrospinal fluid, and AQP4 again assembled more densely around vessels. The authors then repeated the experiment in PS19 mice lacking AQP4. TGN-073 no longer enhanced tracer influx and did not reduce tau pathology, gliosis, or neuronal loss. In this model, the entire observed effect of TGN-073 required AQP4. The connection between TGN-073 and fluid movement did not come out of nowhere: in a 2023 study, MRI also showed wider contrast distribution in healthy rats after this compound. The new article adds a model of tauopathy: here, one target is associated with both fluid movement and tau accumulation and neuronal preservation. For tauopathy, this represents two different directions of intervention: reducing tau production or changing fluid movement, which removes it from tissue. In chimeric mice, AAV vectors have already been used to deliver genetic cargo to human glial cells through cerebrospinal fluid movement. And VY1706 is being tried to reduce tau production in the brain with a single viral injection. In the PS19 model, TGN-073 worked through a different route: through AQP4 and fluid flow around vessels.

🔗 Read original →

ABILITY first recorded human brain signals with its system during an operation. On July 21, Swiss ABILITY Neurotech reported on the first operation in which its system recorded human neural signals. The Munich series involves people who have their brain tumors removed: during the procedure, electrodes will record cortical activity for 20-30 minutes. When a person with paralysis tries to say a word or move their hand, the intention may arise in the cortex, although the signal no longer reaches the muscles and voice. The neurointerface reads the electrical activity of the cortex, and the program learns to associate recurring patterns of activity with commands for a computer, speech synthesizer, or prosthesis. ABILITY's first procedure took place in the neurosurgical department of the Technical University of Munich's university clinic. The first participant was under general anesthesia. The team checked that the electrodes, equipment, and data recording worked together. In subsequent operations, the company plans to offer awake participants speech and motor tasks. Such tasks allow matching the pattern of brain activity with the attempt to utter a word or make a movement. "We can record known phenomena - evoked potentials and high-frequency gamma activity - and compare the quality of the ABILITY signal with clinical electrophysiology systems," says neurosurgeon Simon Jacob. Evoked potential is a brief response of the brain to a stimulus; high-frequency gamma activity is one type of its electrical activity. In the Munich series, the team will compare these signals with recordings from clinical systems. In parallel, in Utrecht, recruitment is underway for the INTRECOM study for people with severe paralysis and communication disorders. There, during the operation, four arrays of electrodes will be placed on the surface of the brain, and then researchers will train participants to use the system at home. The Munich series checks the recording during the operation, while INTRECOM builds a communication channel beyond it.

🔗 Read original →