Muscle Power Measured
A new method measures human muscle strength without implants, using ultra-wideband radar attached to the skin. This is crucial for exoskeletons, robotic prosthetics, and rehabilitation, as these systems must accurately know the force a person is trying to apply. Previously, precise measurements required implanting sensors directly into muscle tissue, making widespread use impossible.
The small radar antenna sends short electromagnetic pulses into the muscle and analyzes the reflected signal. As the muscle contracts, its electromagnetic properties change, and the radar detects these changes, allowing strength assessment without surgery. However, there is a technical drawback - measurements occur with a slight delay, and scientists are working to accelerate processing to real-time for instant exoskeleton response.
In rehabilitation, electrical muscle stimulation is often used, but without precise sensors, it's easy to cause overexertion or tissue damage, as the doctor can't see how hard the muscle is working. Radar sensors can solve this problem, making therapy safer. In sports, such sensors will help determine the exact moment when it's safe to return to training after an injury, as reported in Science Robotics, 2023.
🔗 Read original →
A new method measures human muscle strength without implants, using ultra-wideband radar attached to the skin. This is crucial for exoskeletons, robotic prosthetics, and rehabilitation, as these systems must accurately know the force a person is trying to apply. Previously, precise measurements required implanting sensors directly into muscle tissue, making widespread use impossible.
The small radar antenna sends short electromagnetic pulses into the muscle and analyzes the reflected signal. As the muscle contracts, its electromagnetic properties change, and the radar detects these changes, allowing strength assessment without surgery. However, there is a technical drawback - measurements occur with a slight delay, and scientists are working to accelerate processing to real-time for instant exoskeleton response.
In rehabilitation, electrical muscle stimulation is often used, but without precise sensors, it's easy to cause overexertion or tissue damage, as the doctor can't see how hard the muscle is working. Radar sensors can solve this problem, making therapy safer. In sports, such sensors will help determine the exact moment when it's safe to return to training after an injury, as reported in Science Robotics, 2023.
🔗 Read original →
Science Robotics
Ultra-wideband radar to measure in vivo muscle forces
Ultra-wideband radar scans of skeletal muscle accurately estimate muscle forces during a range of muscle contractions.
Brain Robot Interface
The development by KAIST creates a fully two-way interface between the brain and a robotized exoskeleton. Unlike ordinary BCIs that can only read brain signals and convert them into commands, this development does two things - it receives signals from the brain and sends back sensory information from the robot. This transforms the exoskeleton into not just "external legs", but an extension of the body that a person can feel. The brain chip reads hundreds of channels of cortical activity, AI algorithms interpret movement intention and transmit the command to the exoskeleton.
The robot is equipped with force, moment, and tactile pressure sensors, and this real-time data is encoded into a form that the brain can perceive as sensation. The main complexity is to ensure a stable closed loop, where hundreds of neural channels are processed quickly enough for a person to walk, lift objects, and feel them without delay. The project combines several areas: control of robotized legs, interpretation of movement intentions, creation of "robotic skin" that replaces lost sensitivity, and development of ultra-low-power wireless interfaces for stimulation and recording of neural activity.
The KAIST project involves teams working on miniature neuroelectrodes, neuroengineering, AI chips, and rehabilitation robots to assemble everything into a single architecture, as described in the Journal of Neuroscience Research.
🔗 Source: @solid_state_humanity
The development by KAIST creates a fully two-way interface between the brain and a robotized exoskeleton. Unlike ordinary BCIs that can only read brain signals and convert them into commands, this development does two things - it receives signals from the brain and sends back sensory information from the robot. This transforms the exoskeleton into not just "external legs", but an extension of the body that a person can feel. The brain chip reads hundreds of channels of cortical activity, AI algorithms interpret movement intention and transmit the command to the exoskeleton.
The robot is equipped with force, moment, and tactile pressure sensors, and this real-time data is encoded into a form that the brain can perceive as sensation. The main complexity is to ensure a stable closed loop, where hundreds of neural channels are processed quickly enough for a person to walk, lift objects, and feel them without delay. The project combines several areas: control of robotized legs, interpretation of movement intentions, creation of "robotic skin" that replaces lost sensitivity, and development of ultra-low-power wireless interfaces for stimulation and recording of neural activity.
The KAIST project involves teams working on miniature neuroelectrodes, neuroengineering, AI chips, and rehabilitation robots to assemble everything into a single architecture, as described in the Journal of Neuroscience Research.
🔗 Source: @solid_state_humanity
Telegram
Solid State Humanity
Полностью двунаправленный интерфейс между мозгом и роботизированным экзоскелетом
Если обычные BCI умеют только считывать сигналы мозга и превращать их в команды, то разработка KAIST делает сразу две вещи - принимает сигналы от мозга и отправляет обратно…
Если обычные BCI умеют только считывать сигналы мозга и превращать их в команды, то разработка KAIST делает сразу две вещи - принимает сигналы от мозга и отправляет обратно…
KinoPlex Map Released
The authors of KinoPlex published a map of sites in human proteins where kinases can attach a phosphate on July 29. The researchers connected predicted AlphaFold structures of nearly 20,000 proteins with motifs - short sequence segments that kinases recognize. This map selects positions where a suitable motif matches the available arrangement in the protein's three-dimensional form.
A kinase attaches a phosphate to a protein and changes its function, and through such switches, the cell responds to nutrition, stress, and growth signals. When a researcher looks for a kinase target, they usually look at several amino acids near the supposed site, which together form a motif that fits this enzyme. The KinoPlex authors applied the map to 1.8 million serine, threonine, and tyrosine residues - amino acids to which phosphate is usually attached.
The map marked about 567 thousand positions as accessible for phosphorylation, and matching with individual kinase motifs left about 250 thousand "kinase - position" pairs where both conditions match. The authors verified the predictions with deep phosphoproteomics of K562 cells, a method that massively measures protein segments with attached phosphate. In these measurements, KinoPlex candidates matched the actually observed phosphorylated positions, as reported in Nature Aging, July 2026.
🔗 Read original →
The authors of KinoPlex published a map of sites in human proteins where kinases can attach a phosphate on July 29. The researchers connected predicted AlphaFold structures of nearly 20,000 proteins with motifs - short sequence segments that kinases recognize. This map selects positions where a suitable motif matches the available arrangement in the protein's three-dimensional form.
A kinase attaches a phosphate to a protein and changes its function, and through such switches, the cell responds to nutrition, stress, and growth signals. When a researcher looks for a kinase target, they usually look at several amino acids near the supposed site, which together form a motif that fits this enzyme. The KinoPlex authors applied the map to 1.8 million serine, threonine, and tyrosine residues - amino acids to which phosphate is usually attached.
The map marked about 567 thousand positions as accessible for phosphorylation, and matching with individual kinase motifs left about 250 thousand "kinase - position" pairs where both conditions match. The authors verified the predictions with deep phosphoproteomics of K562 cells, a method that massively measures protein segments with attached phosphate. In these measurements, KinoPlex candidates matched the actually observed phosphorylated positions, as reported in Nature Aging, July 2026.
🔗 Read original →
Nature
An AI-enabled structural atlas decodes kinase specificity across the human proteome
Nature Biotechnology - Phosphorylation potential and kinase specificity are assigned for the entire human proteome.
Bone Marrow Cells
Researchers tracked the origin of immune cells in the brains of 20 elderly donors in a Nature article on July 30. They found descendants of bone marrow cells, similar to microglia, in all samples. Microglia clears cellular debris and responds to damage. In mouse experiments, it colonizes the brain before birth and then multiplies locally.
The team of Julia Belk checked if this is also the case in the aging human brain. Over time, dividing cells acquire random DNA mutations that are inherited by their descendants. The researchers used these somatic mutations as markers of kinship, comparing them in blood, bone marrow, and brain cells. Matches showed that several myeloid lines came from the bone marrow to the brain. Single-cell analysis and mitochondrial DNA variants revealed similarities between these cells and microglia.
In some samples, they made up a large proportion of the microglia pool. In the oldest donors, the larger the clone in the blood, the more of its descendants were found in the brain. The authors propose a model where, with age, blood cells replace some of the embryonically derived microglia. In this model, the state of the bone marrow changes the brain's immune environment through the composition of cells that enter it. The connection to Alzheimer's disease looks different in different studies. The Nature authors found an association between most types of clonal hematopoiesis and a lower frequency of the disease in human cohorts.
🔗 Read original →
Researchers tracked the origin of immune cells in the brains of 20 elderly donors in a Nature article on July 30. They found descendants of bone marrow cells, similar to microglia, in all samples. Microglia clears cellular debris and responds to damage. In mouse experiments, it colonizes the brain before birth and then multiplies locally.
The team of Julia Belk checked if this is also the case in the aging human brain. Over time, dividing cells acquire random DNA mutations that are inherited by their descendants. The researchers used these somatic mutations as markers of kinship, comparing them in blood, bone marrow, and brain cells. Matches showed that several myeloid lines came from the bone marrow to the brain. Single-cell analysis and mitochondrial DNA variants revealed similarities between these cells and microglia.
In some samples, they made up a large proportion of the microglia pool. In the oldest donors, the larger the clone in the blood, the more of its descendants were found in the brain. The authors propose a model where, with age, blood cells replace some of the embryonically derived microglia. In this model, the state of the bone marrow changes the brain's immune environment through the composition of cells that enter it. The connection to Alzheimer's disease looks different in different studies. The Nature authors found an association between most types of clonal hematopoiesis and a lower frequency of the disease in human cohorts.
🔗 Read original →
Nature
Somatic mutations reveal the ontogeny of microglia in human aging
Nature - Somatic mutations reveal the ontogeny of microglia in human aging
Cell Cryopreservation
The installation converts a cell suspension into droplets and vitrifies it with a flow rate of 100 ml per hour. On July 27, a group from the University of Minnesota released a preprint on cryo-aerosolization: a vibrating nozzle breaks the cell suspension into microdroplets, and a jet of liquid nitrogen freezes them. After thawing, about 90% of human fibroblasts and induced pluripotent stem cells remained viable.
The cells for therapy are first grown, then stored and transported. When frozen, water forms ice crystals that damage cell membranes and internal structures. Vitrification protects cells from ice if the solution is rapidly cooled and rapidly thawed. With small volumes, this is easier to achieve. A microdroplet quickly releases heat: its surface is large relative to its volume. A cell dose takes tens or hundreds of milliliters, and a large portion cools more slowly.
In the installation, the nozzle creates droplets with a diameter of about 200 micrometers. The jet of liquid nitrogen collides with them in flight, and the droplets fall into a collector for storage. According to the authors' thermal model, the collision with nitrogen accelerates heat removal and does not allow the droplet to be suspended over nitrogen on a vapor layer. In the experiment, 100 ml of cell suspension passed through the installation per hour; the proportion of penetrating cryoprotector was 19-25%. Instead of one large portion, the installation freezes multiple microdroplets. The authors measured an average cooling rate of up to 210,000 degrees per minute, and their thawing model gave about 1 million degrees per minute. After one cycle, fibroblasts and induced pluripotent cells retained viability of about 90%, pig erythrocytes recovered at 94%. For induced pluripotent cells, the team also checked colony formation after re-seeding, as described in the review of cryopreservation of cell therapies in Nature Aging, July 2026.
🔗 Read original →
The installation converts a cell suspension into droplets and vitrifies it with a flow rate of 100 ml per hour. On July 27, a group from the University of Minnesota released a preprint on cryo-aerosolization: a vibrating nozzle breaks the cell suspension into microdroplets, and a jet of liquid nitrogen freezes them. After thawing, about 90% of human fibroblasts and induced pluripotent stem cells remained viable.
The cells for therapy are first grown, then stored and transported. When frozen, water forms ice crystals that damage cell membranes and internal structures. Vitrification protects cells from ice if the solution is rapidly cooled and rapidly thawed. With small volumes, this is easier to achieve. A microdroplet quickly releases heat: its surface is large relative to its volume. A cell dose takes tens or hundreds of milliliters, and a large portion cools more slowly.
In the installation, the nozzle creates droplets with a diameter of about 200 micrometers. The jet of liquid nitrogen collides with them in flight, and the droplets fall into a collector for storage. According to the authors' thermal model, the collision with nitrogen accelerates heat removal and does not allow the droplet to be suspended over nitrogen on a vapor layer. In the experiment, 100 ml of cell suspension passed through the installation per hour; the proportion of penetrating cryoprotector was 19-25%. Instead of one large portion, the installation freezes multiple microdroplets. The authors measured an average cooling rate of up to 210,000 degrees per minute, and their thawing model gave about 1 million degrees per minute. After one cycle, fibroblasts and induced pluripotent cells retained viability of about 90%, pig erythrocytes recovered at 94%. For induced pluripotent cells, the team also checked colony formation after re-seeding, as described in the review of cryopreservation of cell therapies in Nature Aging, July 2026.
🔗 Read original →
bioRxiv
Cryoaerosolization Enables Scalable Vitrification-Based Cell Cryopreservation
Cell therapies hold transformative potential for treating cancer, neurologic disorders, organ failure, diabetes, and other conditions, but their widespread clinical deployment is constrained by the lack of scalable cryopreservation methods that maintain high…
Tomorrow Bio Training
Tomorrow Bio demonstrated a training rehearsal for cryopreservation on August 7. The company released a video showing the team's training on a specially made mannequin, from the initial cooling to the introduction of a cryoprotector. The training begins before the operation, with the team loading equipment into a vehicle and transporting the mannequin on a stretcher with an ice bath.
The team then initiates cooling, chest compressions, oxygen supply, and medication administration. Once the temperature is lowered, the surgeon has time for the next step, which involves opening the chest cavity and inserting a cannula into the aorta. A perfusion solution is then administered through the cannula, followed by the cryoprotector solution, which allows tissues to be cooled to extremely low temperatures without forming ice crystals.
The team measures pressure, temperature, and refractive index as the solution passes through the vessels, helping to understand the concentration of the cryoprotector. The procedure is recorded by cameras for future training, and a remote physician can connect through one of the cameras. After perfusion, the scenario involves continued cooling and transportation to Switzerland, where a computer tomography scan and electron microscopy analysis of microsamples would be conducted, as described in a previous explanation by Tomorrow Bio in March.
🔗 Read original →
Tomorrow Bio demonstrated a training rehearsal for cryopreservation on August 7. The company released a video showing the team's training on a specially made mannequin, from the initial cooling to the introduction of a cryoprotector. The training begins before the operation, with the team loading equipment into a vehicle and transporting the mannequin on a stretcher with an ice bath.
The team then initiates cooling, chest compressions, oxygen supply, and medication administration. Once the temperature is lowered, the surgeon has time for the next step, which involves opening the chest cavity and inserting a cannula into the aorta. A perfusion solution is then administered through the cannula, followed by the cryoprotector solution, which allows tissues to be cooled to extremely low temperatures without forming ice crystals.
The team measures pressure, temperature, and refractive index as the solution passes through the vessels, helping to understand the concentration of the cryoprotector. The procedure is recorded by cameras for future training, and a remote physician can connect through one of the cameras. After perfusion, the scenario involves continued cooling and transportation to Switzerland, where a computer tomography scan and electron microscopy analysis of microsamples would be conducted, as described in a previous explanation by Tomorrow Bio in March.
🔗 Read original →
www.tomorrow.bio
Engineering realism: How advanced training dummies are transforming Cryonics readiness
Tomorrow.bio develops advanced training dummies for cryonics preparedness, including Weighted, SST, and Surgical Dummies, enhancing realism and precision for critical procedures.
Old Mice Regain Strength
Researchers introduced SW033291, a substance that blocks 15-PGDH, to 28-month-old mice for two weeks, and found that the strength of their foot flexors increased by approximately 50% compared to control animals. The study, published in preprint, August 7, linked the result to the growth signal IGF1.
As mice age, their muscles become less responsive to exercise, a phenomenon known as anabolic resistance. When muscles are overloaded, muscle fibers and surrounding cells exchange signals to initiate repair and growth. The authors surgically overloaded one muscle in the calf and then administered SW033291 daily to the mice.
The treatment increased PGE2 levels in the overloaded muscle, as well as its mass and fiber area. The strength of the foot flexors was measured by electrical nerve stimulation and found to be approximately 50% higher in mice receiving SW033291 compared to those receiving a solvent instead.
To understand the mechanism, researchers analyzed RNA from approximately 200,000 cell nuclei in muscle tissue. The analysis showed that blocking 15-PGDH increased the activity of the Igf1 gene, which codes for the protein growth factor IGF1, in one type of fast muscle fiber. A computer model predicted that IGF1 may transmit signals between several cell types in the muscle.
🔗 Read original →
Researchers introduced SW033291, a substance that blocks 15-PGDH, to 28-month-old mice for two weeks, and found that the strength of their foot flexors increased by approximately 50% compared to control animals. The study, published in preprint, August 7, linked the result to the growth signal IGF1.
As mice age, their muscles become less responsive to exercise, a phenomenon known as anabolic resistance. When muscles are overloaded, muscle fibers and surrounding cells exchange signals to initiate repair and growth. The authors surgically overloaded one muscle in the calf and then administered SW033291 daily to the mice.
The treatment increased PGE2 levels in the overloaded muscle, as well as its mass and fiber area. The strength of the foot flexors was measured by electrical nerve stimulation and found to be approximately 50% higher in mice receiving SW033291 compared to those receiving a solvent instead.
To understand the mechanism, researchers analyzed RNA from approximately 200,000 cell nuclei in muscle tissue. The analysis showed that blocking 15-PGDH increased the activity of the Igf1 gene, which codes for the protein growth factor IGF1, in one type of fast muscle fiber. A computer model predicted that IGF1 may transmit signals between several cell types in the muscle.
🔗 Read original →
Lifespan Simulator
The Lifespan Simulator, published by Александр Фединцев, calculates the proportion of people who will live to a chosen age. Users can input the timeline for preventing different causes of death, and the model calculates the percentage of the current population that will reach the selected age.
The simulator takes into account конкурирующие причины смерти, where one disease claims a person's life before another can. For example, if cancer is cured, the person may still die from болезни сердца и сосудов or респираторных болезней. In 1999, epidemiologists J. P. Mackenbach, Anton Kunst, and colleagues analyzed data from 5,975 Dutch death certificates from 1990, showing that eliminating one cause of death changes the comparison of gains from eliminating others.
The Lifespan Simulator tracks the fate of a single cohort year by year, allowing users to input the timeline for preventing рак, деменции, and болезни сосудов, as well as the age to which they want to calculate the proportion of survivors. The result is the proportion of people from the same cohort who will live to that age, as reported in Nature Aging, July 2026.
🔗 Read original →
The Lifespan Simulator, published by Александр Фединцев, calculates the proportion of people who will live to a chosen age. Users can input the timeline for preventing different causes of death, and the model calculates the percentage of the current population that will reach the selected age.
The simulator takes into account конкурирующие причины смерти, where one disease claims a person's life before another can. For example, if cancer is cured, the person may still die from болезни сердца и сосудов or респираторных болезней. In 1999, epidemiologists J. P. Mackenbach, Anton Kunst, and colleagues analyzed data from 5,975 Dutch death certificates from 1990, showing that eliminating one cause of death changes the comparison of gains from eliminating others.
The Lifespan Simulator tracks the fate of a single cohort year by year, allowing users to input the timeline for preventing рак, деменции, and болезни сосудов, as well as the age to which they want to calculate the proportion of survivors. The result is the proportion of people from the same cohort who will live to that age, as reported in Nature Aging, July 2026.
🔗 Read original →
Demographic Research
Demographic Research - An integrated approach to cause-of-death analysis: cause-deleted life tables and decompositions of life…
Longevity Channels
For those interested in life extension, immortality, and transhumanism, a thematic folder of channels has been compiled at https://t.me/addlist/ywXQpGsp50wxYzNi. You can access the collection if the topic is of interest to you.
The folder includes a range of transhumanism topics, from theoretical discussions to practical applications. It is intended for individuals who want to explore the possibilities of life extension and immortality.
The collection is available for anyone to access, with a focus on providing resources and information on biotech advancements and their potential impact on human longevity.
🔗 Source: @solid_state_humanity
For those interested in life extension, immortality, and transhumanism, a thematic folder of channels has been compiled at https://t.me/addlist/ywXQpGsp50wxYzNi. You can access the collection if the topic is of interest to you.
The folder includes a range of transhumanism topics, from theoretical discussions to practical applications. It is intended for individuals who want to explore the possibilities of life extension and immortality.
The collection is available for anyone to access, with a focus on providing resources and information on biotech advancements and their potential impact on human longevity.
🔗 Source: @solid_state_humanity
Telegram
Имморталисты
nuM1nex invites you to add the folder “Имморталисты”, which includes 20 chats.
Longevity Channels
For those interested in life extension, immortality, and transhumanism, a thematic folder of channels has been compiled at https://t.me/addlist/ywXQpGsp50wxYzNi. You can access the collection if the topic is of interest to you.
The collection is available for those who want to explore transhumanism and related topics, including immortality and life extension. It can be found through the provided link.
The folder contains a list of channels that discuss various aspects of life extension, making it a valuable resource for those interested in this field, with references to studies published in Nature Aging, July 2026.
🔗 Source: @solid_state_humanity
For those interested in life extension, immortality, and transhumanism, a thematic folder of channels has been compiled at https://t.me/addlist/ywXQpGsp50wxYzNi. You can access the collection if the topic is of interest to you.
The collection is available for those who want to explore transhumanism and related topics, including immortality and life extension. It can be found through the provided link.
The folder contains a list of channels that discuss various aspects of life extension, making it a valuable resource for those interested in this field, with references to studies published in Nature Aging, July 2026.
🔗 Source: @solid_state_humanity
Telegram
Имморталисты
nuM1nex invites you to add the folder “Имморталисты”, which includes 20 chats.
Bionic Eyes Advance
In an experiment, scientists used a deep neural network to control electrical stimulation in the visual cortex, the brain area responsible for processing visual information. A blind participant with a temporary implant of 96 electrodes in the cortex saw phosphenes, or flashes of light, when the electrodes were stimulated. The main goal was to accurately predict which phosphenes would appear with specific stimulation and make these sensations more controllable.
The neural network was trained on real brain activity data, taking into account the state of the brain before each test, including its "background" activity. This allowed the stimulation to be adapted to the current neural context. When researchers used stimulation patterns selected by AI, they achieved more accurate reproduction of the desired brain activity with less electrical current.
The brain activity proved to be a better predictor of what the person would see than the stimulation parameters themselves. This is crucial because the brain does not work like a screen with pixels, and stimulation of one electrode affects neighboring ones, with neural responses changing from day to day. Therefore, future bionic eyes should be adaptive, with the device adjusting to the brain, not the other way around, as reported in Nature Aging, July 2026.
🔗 Read original →
In an experiment, scientists used a deep neural network to control electrical stimulation in the visual cortex, the brain area responsible for processing visual information. A blind participant with a temporary implant of 96 electrodes in the cortex saw phosphenes, or flashes of light, when the electrodes were stimulated. The main goal was to accurately predict which phosphenes would appear with specific stimulation and make these sensations more controllable.
The neural network was trained on real brain activity data, taking into account the state of the brain before each test, including its "background" activity. This allowed the stimulation to be adapted to the current neural context. When researchers used stimulation patterns selected by AI, they achieved more accurate reproduction of the desired brain activity with less electrical current.
The brain activity proved to be a better predictor of what the person would see than the stimulation parameters themselves. This is crucial because the brain does not work like a screen with pixels, and stimulation of one electrode affects neighboring ones, with neural responses changing from day to day. Therefore, future bionic eyes should be adaptive, with the device adjusting to the brain, not the other way around, as reported in Nature Aging, July 2026.
🔗 Read original →
Neuron
Deep learning-based control of electrically evoked activity in human visual cortex
We show that a deep learning framework can predict and shape stimulation-evoked intracortical
neural activity in the visual cortex of a blind human participant. Optimized stimulation
patterns accurately produce target neural responses at low currents, while…
neural activity in the visual cortex of a blind human participant. Optimized stimulation
patterns accurately produce target neural responses at low currents, while…
NAMPT Activator Slows Frailty
The activator of the NAMPT enzyme, combined with nicotinamide, slowed the progression of frailty in old mice. In a preprint from August 7, a team from Sanford Burnham Prebys described the compound SBI-0802162, which activates the NAMPT enzyme. In human cell culture, it more strongly increased NAD+ levels in senescent cells than in dividing cells, and with prolonged exposure, reduced their viability.
In 18-month-old mice, after 12 weeks of combination with nicotinamide, the frailty index remained stable; in the rotating rod test, the authors saw a strong trend towards better performance than in the control. Senescent cells stop dividing after damage or stress, and the authors sought to find a vulnerability in them through NAD+ metabolism - a molecule involved in energy production, DNA repair, and cell stress response.
When NAD+ is depleted, nicotinamide remains; the NAMPT enzyme triggers its recycling back into NAD+. Isotopic labeling allowed the authors to distinguish between the total NAD+ pool and the rate of its synthesis and consumption. In human connective tissue cell culture, translated into a senescent state by radiation, NAMPT was abundant, although NAD+ was formed and consumed more slowly, indicating that part of NAMPT activity could remain unused. The SBI-0802162 compound activates NAMPT.
For more information, see the study published in Nature Aging, July 2026.
🔗 Read original →
The activator of the NAMPT enzyme, combined with nicotinamide, slowed the progression of frailty in old mice. In a preprint from August 7, a team from Sanford Burnham Prebys described the compound SBI-0802162, which activates the NAMPT enzyme. In human cell culture, it more strongly increased NAD+ levels in senescent cells than in dividing cells, and with prolonged exposure, reduced their viability.
In 18-month-old mice, after 12 weeks of combination with nicotinamide, the frailty index remained stable; in the rotating rod test, the authors saw a strong trend towards better performance than in the control. Senescent cells stop dividing after damage or stress, and the authors sought to find a vulnerability in them through NAD+ metabolism - a molecule involved in energy production, DNA repair, and cell stress response.
When NAD+ is depleted, nicotinamide remains; the NAMPT enzyme triggers its recycling back into NAD+. Isotopic labeling allowed the authors to distinguish between the total NAD+ pool and the rate of its synthesis and consumption. In human connective tissue cell culture, translated into a senescent state by radiation, NAMPT was abundant, although NAD+ was formed and consumed more slowly, indicating that part of NAMPT activity could remain unused. The SBI-0802162 compound activates NAMPT.
For more information, see the study published in Nature Aging, July 2026.
🔗 Read original →
bioRxiv
NAMPT activation uncovers a senescence-specific vulnerability and promotes healthy aging in combination with NAM
Aging is driven by multiple interacting processes, suggesting that effective strategies to promote healthy aging may require simultaneous targeting of more than one underlying mechanism. Here we identify a strategy that couples restoration of nicotinamide…
Lifespan.com Launched
Biologist David Sinclair launched Lifespan.com on August 7, a platform that connects scientific media, community, and research support for aging. The new organization combines Lifespan Magazine, Sinclair's show, educational materials, meetings with scientists, and a membership community, while also supporting the Lifespan Foundation, which aids medical research and aging studies.
Lifespan.com offers readers a consistent route: they can read research analyses in Lifespan Magazine, listen to scientists and ask them questions on the show and at meetings, then join the community and support research through the foundation. The magazine sets the topics for discussion, meetings connect the audience with researchers, and the foundation gathers support for new works.
The magazine starts this chain: the reader first receives a research analysis, then decides which questions to ask the scientists and which works to support. Sinclair explained the reason for the launch as follows: "We are entering a period when discoveries in the biology of aging are coming out of the lab and into everyday life, but public understanding is not keeping up." In the founding article, the magazine team promises to indicate the sources of scientific claims and evaluate advice for their validity, benefits, and connection to data, as published in Lifespan Magazine.
🔗 Read original →
Biologist David Sinclair launched Lifespan.com on August 7, a platform that connects scientific media, community, and research support for aging. The new organization combines Lifespan Magazine, Sinclair's show, educational materials, meetings with scientists, and a membership community, while also supporting the Lifespan Foundation, which aids medical research and aging studies.
Lifespan.com offers readers a consistent route: they can read research analyses in Lifespan Magazine, listen to scientists and ask them questions on the show and at meetings, then join the community and support research through the foundation. The magazine sets the topics for discussion, meetings connect the audience with researchers, and the foundation gathers support for new works.
The magazine starts this chain: the reader first receives a research analysis, then decides which questions to ask the scientists and which works to support. Sinclair explained the reason for the launch as follows: "We are entering a period when discoveries in the biology of aging are coming out of the lab and into everyday life, but public understanding is not keeping up." In the founding article, the magazine team promises to indicate the sources of scientific claims and evaluate advice for their validity, benefits, and connection to data, as published in Lifespan Magazine.
🔗 Read original →
PR Newswire
Dr. David Sinclair Launches Lifespan, the First Science Media Platform Built by Longevity Scientists to Advance Medical Research
/PRNewswire/ -- Dr. David A. Sinclair, A.O., Ph.D., Professor of Genetics at Harvard Medical School, longevity scientist, and author of the international...
AI Designs Viruses
The recent work where "AI created a virus" has been surrounded by controversy, but the reality is much more subdued. Yes, AI has indeed learned to design entire bacteriophages, but not for apocalyptic purposes, rather for treating infections that no longer respond to antibiotics. Bacteriophages are not separate genes or small DNA fragments, but complete genomes approximately 5,300 nucleotides in length, with their own structure, regulation, and set of proteins.
The study used two language-based genomic AI models, Evo 1 and Evo 2, which generated thousands of phage variants, after which researchers chemically synthesized almost 300 of them and tested them in the laboratory. As a result, 16 fully viable phages were obtained that infected Escherichia coli, had different replication rates, different structures, and even used proteins not found in the natural prototype φX174. One of the synthetic phages incorporated a DNA packaging protein from an evolutionarily distant virus into its capsid, meaning the AI created a combination that does not occur in nature but works.
The most impressive aspect is that a mixture of AI-created phages was able to quickly destroy E. coli strains resistant to the natural φX174, while a mixture of natural phages, even those with similar structures, was unable to accomplish the same task. Generative models can design phages that bypass bacterial resistance, change their infection strategy, and form new evolutionary combinations inaccessible to conventional bioengineering. In essence, this is the first step towards creating phage therapies for specific resistant infections - quickly, precisely, and with specified properties.
🔗 Read original →
The recent work where "AI created a virus" has been surrounded by controversy, but the reality is much more subdued. Yes, AI has indeed learned to design entire bacteriophages, but not for apocalyptic purposes, rather for treating infections that no longer respond to antibiotics. Bacteriophages are not separate genes or small DNA fragments, but complete genomes approximately 5,300 nucleotides in length, with their own structure, regulation, and set of proteins.
The study used two language-based genomic AI models, Evo 1 and Evo 2, which generated thousands of phage variants, after which researchers chemically synthesized almost 300 of them and tested them in the laboratory. As a result, 16 fully viable phages were obtained that infected Escherichia coli, had different replication rates, different structures, and even used proteins not found in the natural prototype φX174. One of the synthetic phages incorporated a DNA packaging protein from an evolutionarily distant virus into its capsid, meaning the AI created a combination that does not occur in nature but works.
The most impressive aspect is that a mixture of AI-created phages was able to quickly destroy E. coli strains resistant to the natural φX174, while a mixture of natural phages, even those with similar structures, was unable to accomplish the same task. Generative models can design phages that bypass bacterial resistance, change their infection strategy, and form new evolutionary combinations inaccessible to conventional bioengineering. In essence, this is the first step towards creating phage therapies for specific resistant infections - quickly, precisely, and with specified properties.
🔗 Read original →
Science
Generative design of bacteriophages with genome language models
Many important biological functions arise not from single genes but from complex interactions encoded by entire genomes. We report the first generative design of complete bacteriophage genomes using genome language models. We generated viable ...
Conduit Brain Signal Collection
Keller Scholl published an essay on Conduit, a company that collects non-invasive neurodata, or brain signals from the surface of the head, to train models to translate them into text. He calls for stopping this work at an early stage, including data collection, device development, and funding.
The Conduit project page describes a two-hour conversation between a participant and a language model in a headset, during which the team records brain signals, text, and audio. In December, the company reported approximately 10,000 hours of such recordings from thousands of participants. Each recording gives the model a pair for training: a brain signal and a phrase spoken or typed by the participant.
In Nature Aging, July 2026, a team led by Jerry Tang restored the meaning of perceived and imagined speech using functional MRI, which shows brain activity. This technology can restore communication to people who have lost normal speech or movement. Scholl suggests addressing the fate of such data before they enter a large archive, and warns against the potential for forced application of this technology.
🔗 Read original →
Keller Scholl published an essay on Conduit, a company that collects non-invasive neurodata, or brain signals from the surface of the head, to train models to translate them into text. He calls for stopping this work at an early stage, including data collection, device development, and funding.
The Conduit project page describes a two-hour conversation between a participant and a language model in a headset, during which the team records brain signals, text, and audio. In December, the company reported approximately 10,000 hours of such recordings from thousands of participants. Each recording gives the model a pair for training: a brain signal and a phrase spoken or typed by the participant.
In Nature Aging, July 2026, a team led by Jerry Tang restored the meaning of perceived and imagined speech using functional MRI, which shows brain activity. This technology can restore communication to people who have lost normal speech or movement. Scholl suggests addressing the fate of such data before they enter a large archive, and warns against the potential for forced application of this technology.
🔗 Read original →
PubMed Central (PMC)
Semantic reconstruction of continuous language from non-invasive brain recordings
A brain-computer interface that decodes continuous language from non-invasive recordings would have many scientific and practical applications. Currently, however, non-invasive language decoders can only identify stimuli from among a small set of ...
VirTues Model
The VirTues model compares tissue snapshots with different sets of proteins. On August 5, an article about VirTues was published in Nature: it is a model for spatial proteomics that analyzes proteins in a tissue section along with their arrangement. It was trained on 32 clinical cohorts, data from more than 5,100 patients, and 239 proteins.
The authors tested the model on biopsies of patients with triple-negative breast cancer. In the tumor, cancerous, immune, and connective tissue cells coexist. Spatial proteomics turns a tissue section into a map: it shows which proteins are present in cells and which cells are neighboring each other. Two samples may contain the same types of cells but differ in their arrangement.
The VirTues model associates the signal of each measured protein with its amino acid sequence and its arrangement in the tissue section. During training, the model hides part of the signal and reconstructs it based on neighboring proteins and the layout of the area. This allows it to learn to consider both the individual cell and its environment.
In a strict test, the authors completely excluded the target cohort from training and then asked VirTues to identify cells in its snapshots. The authors counted a cell as correctly identified if the predicted boundary matched the annotation. By this criterion, the system outperformed three specialized programs on eight out of nine datasets. On biopsies of 111 patients with triple-negative breast cancer, taken before treatment, VirTues identified four spatial signatures - combinations of cell types and their co-occurrence.
🔗 Read original →
The VirTues model compares tissue snapshots with different sets of proteins. On August 5, an article about VirTues was published in Nature: it is a model for spatial proteomics that analyzes proteins in a tissue section along with their arrangement. It was trained on 32 clinical cohorts, data from more than 5,100 patients, and 239 proteins.
The authors tested the model on biopsies of patients with triple-negative breast cancer. In the tumor, cancerous, immune, and connective tissue cells coexist. Spatial proteomics turns a tissue section into a map: it shows which proteins are present in cells and which cells are neighboring each other. Two samples may contain the same types of cells but differ in their arrangement.
The VirTues model associates the signal of each measured protein with its amino acid sequence and its arrangement in the tissue section. During training, the model hides part of the signal and reconstructs it based on neighboring proteins and the layout of the area. This allows it to learn to consider both the individual cell and its environment.
In a strict test, the authors completely excluded the target cohort from training and then asked VirTues to identify cells in its snapshots. The authors counted a cell as correctly identified if the predicted boundary matched the annotation. By this criterion, the system outperformed three specialized programs on eight out of nine datasets. On biopsies of 111 patients with triple-negative breast cancer, taken before treatment, VirTues identified four spatial signatures - combinations of cell types and their co-occurrence.
🔗 Read original →
Nature
The Virtual Tissues foundation model resolves spatial proteomics across scales
Nature - Virtual Tissues (VirTues), a foundation model for spatial proteomics that captures tissue organization across scales, supports marker reconstruction, cell segmentation and typing, niche...
Reviving Old T-Cells
The human body's ability to produce new T-cells declines with age, and existing T-cells become slower, less able to divide, and less effective at recognizing threats. A team of scientists sought to determine if they could restore some of the lost function in these cells by reprogramming them from the inside. The main technical challenge was delivering new instructions to old T-cells without damaging them.
The researchers used silicon nanowires - microscopic structures that allow molecular signals to be introduced directly into a cell. With this method, they were able to reprogram more than 90 percent of old T-cells without causing damage. After reprogramming, the cells became more active, dividing more quickly and better attacking infected and cancerous cells, behaving almost like young cells. The team was surprised to find that changing the function of only 4-5 key genes was enough to restore function.
The effect was tested on T-cells from elderly individuals, cancer patients, and cancer survivors, and in all groups, the cells became noticeably more active. The effect currently lasts for around two weeks, but the team is working to extend it, as reported in Nature Aging, July 2026. This does not reverse biological aging, but allows cells to temporarily stop "behaving like old cells" and resume their protective functions.
🔗 Read original →
The human body's ability to produce new T-cells declines with age, and existing T-cells become slower, less able to divide, and less effective at recognizing threats. A team of scientists sought to determine if they could restore some of the lost function in these cells by reprogramming them from the inside. The main technical challenge was delivering new instructions to old T-cells without damaging them.
The researchers used silicon nanowires - microscopic structures that allow molecular signals to be introduced directly into a cell. With this method, they were able to reprogram more than 90 percent of old T-cells without causing damage. After reprogramming, the cells became more active, dividing more quickly and better attacking infected and cancerous cells, behaving almost like young cells. The team was surprised to find that changing the function of only 4-5 key genes was enough to restore function.
The effect was tested on T-cells from elderly individuals, cancer patients, and cancer survivors, and in all groups, the cells became noticeably more active. The effect currently lasts for around two weeks, but the team is working to extend it, as reported in Nature Aging, July 2026. This does not reverse biological aging, but allows cells to temporarily stop "behaving like old cells" and resume their protective functions.
🔗 Read original →
Cell Biomaterials
miRNA delivery via nanowires restores functional responses in aged T cells
Aging reshapes CD8+ T cells, reducing their ability to activate, expand, and respond
to infection or vaccination. Singh and colleagues use biomaterial-functionalized silicon
nanowires to deliver microRNAs into aged T cells with high efficiency and viability.…
to infection or vaccination. Singh and colleagues use biomaterial-functionalized silicon
nanowires to deliver microRNAs into aged T cells with high efficiency and viability.…
Cryopreservation Standards
Maks Mor suggests evaluating cryocenters by their ability to store patients for decades. On August 5, Mor published an essay on biostasis, the preservation of humans after legal death for potential future restoration. He proposes evaluating a cryocenter's work in the first hours and its ability to store a patient for decades.
Mor adds a second criterion to the proposal by Jessica Radley and Ashwin de Wolf to evaluate biostasis methods by neuronal structure preservation, which is the basis of memory and habits. A well-preserved brain will not wait for future medicine if the organization responsible for it does not survive the term. In the first hours, the standby team prepares the body for storage and transports the patient.
Then, the organization buys liquid nitrogen, keeps documents, executes contracts, and passes on this responsibility to people who do not yet work in the organization. Mor formulates the risk directly: "You may be preserved in excellent condition, but you will not see the future because the organization will eventually collapse." The preservation of the body depends on the technique, while the money, documents, and people responsible for storage determine whether the patient will see future medicine, as described in the review of early cryonics by R. Michael Perry.
🔗 Read original →
Maks Mor suggests evaluating cryocenters by their ability to store patients for decades. On August 5, Mor published an essay on biostasis, the preservation of humans after legal death for potential future restoration. He proposes evaluating a cryocenter's work in the first hours and its ability to store a patient for decades.
Mor adds a second criterion to the proposal by Jessica Radley and Ashwin de Wolf to evaluate biostasis methods by neuronal structure preservation, which is the basis of memory and habits. A well-preserved brain will not wait for future medicine if the organization responsible for it does not survive the term. In the first hours, the standby team prepares the body for storage and transports the patient.
Then, the organization buys liquid nitrogen, keeps documents, executes contracts, and passes on this responsibility to people who do not yet work in the organization. Mor formulates the risk directly: "You may be preserved in excellent condition, but you will not see the future because the organization will eventually collapse." The preservation of the body depends on the technique, while the money, documents, and people responsible for storage determine whether the patient will see future medicine, as described in the review of early cryonics by R. Michael Perry.
🔗 Read original →
Substack
Building Biostasis Organizations to Last
No one can predict with any confidence how long it will be before it may be possible to repair and revive patients in biostasis. It is plausible that it will take a century. It could be decades less – especially if artificial intelligence accelerates biomedical…
Colorectal Adenoma Map
Researchers have linked the GDF15 protein to tissue-renewing cells and lower CD8 cell density in a study published on August 7 in a preprint. The authors analyzed spatial measurements of RNA and proteins in 22 adenomas and compared them to 101 fragments of normal tissue, adenomas, and carcinomas from 16 patients.
The study used spatial multi-omics to examine tissue sections while preserving cell arrangement, allowing researchers to see which genes and proteins are active in each area. The authors looked for senescence, a state of prolonged cell division arrest, in specific areas of the polyp and found that these areas were neighboring two groups of altered cells: those with a senescence program and those with a tissue-renewing program.
The authors compared advanced and early adenomas, areas with high and low stemness, as well as senescent areas of adenomas and normal epithelium, and found six proteins that cells secrete outward. They chose GDF15, a stress response protein, as a candidate for local mediator, and found that when they statistically accounted for GDF15, the connection between senescence and stemness in advanced adenomas weakened.
Around areas with high GDF15 levels, CD8 cells, immune cells that can destroy cells with tumor characteristics, were less common. This local pattern was observed in both protein and RNA markers in human adenomas, continuing a research line on GDF15 that started in 2019 with a study showing that senescent fibroblasts, cells of connective tissue, secrete GDF15 and support the growth of adenoma cells in organoids and cell systems.
🔗 Read original →
Researchers have linked the GDF15 protein to tissue-renewing cells and lower CD8 cell density in a study published on August 7 in a preprint. The authors analyzed spatial measurements of RNA and proteins in 22 adenomas and compared them to 101 fragments of normal tissue, adenomas, and carcinomas from 16 patients.
The study used spatial multi-omics to examine tissue sections while preserving cell arrangement, allowing researchers to see which genes and proteins are active in each area. The authors looked for senescence, a state of prolonged cell division arrest, in specific areas of the polyp and found that these areas were neighboring two groups of altered cells: those with a senescence program and those with a tissue-renewing program.
The authors compared advanced and early adenomas, areas with high and low stemness, as well as senescent areas of adenomas and normal epithelium, and found six proteins that cells secrete outward. They chose GDF15, a stress response protein, as a candidate for local mediator, and found that when they statistically accounted for GDF15, the connection between senescence and stemness in advanced adenomas weakened.
Around areas with high GDF15 levels, CD8 cells, immune cells that can destroy cells with tumor characteristics, were less common. This local pattern was observed in both protein and RNA markers in human adenomas, continuing a research line on GDF15 that started in 2019 with a study showing that senescent fibroblasts, cells of connective tissue, secrete GDF15 and support the growth of adenoma cells in organoids and cell systems.
🔗 Read original →
bioRxiv
Spatial multi-omics and single-cell transcriptomics uncover senescence-associated cellular programs during colon adenoma to cancer…
Colorectal cancer develops through a normal-adenoma-carcinoma sequence, yet only 5-10% of adenomas progress to malignancy, and the cellular programs governing that sequence remain poorly defined. Here we generate a spatial multi-omics atlas of human colon…
ARPA-H DNA Project
The American agency ARPA-H has commissioned GE HealthCare to assemble a system for on-demand DNA production and verification. On August 4, DNA Script, a developer of enzyme-based DNA synthesis technology, announced its participation in the four-year FLASH project. ARPA-H has contracted with primary contractor GE HealthCare for up to $26 million; the agency's card indicates the project began on March 16.
The team is to assemble an automated system for DNA manufacturing, purification, and verification, which will be transferred to the program's early users. A researcher starts with a digital sequence, but for the experience, they need a physical DNA molecule. It needs to be manufactured, purified from byproducts, and verified to match the ordered sequence. According to ARPA-H's card, the FLASH project is described as an automated system that should connect these operations.
The first step in this path is DNA Script's enzyme-based synthesis. Enzymes are proteins that initiate chemical reactions; in the company's technology, one such enzyme adds one nucleotide, the DNA building block, to the growing chain at a time. A temporary blocking group on the added nucleotide stops the chain elongation. It is removed, and the enzyme adds the next link. The sequence is assembled one nucleotide per cycle. In the project, DNA Script is responsible for enzyme-based synthesis, while GE HealthCare adds scaling technology to increase production volume.
The program should connect synthesis with purification, verification of the finished DNA's match to the ordered sequence, and automation. In June, synthetic DNA suppliers were called to verify both the ordered sequence and the buyer. For FLASH, ARPA-H requires biosecurity and cybersecurity in addition to production stages. According to the agency's plan, after development, the system will be transferred to the program's early users: researchers working on DNA-based medicines will be able to produce the necessary sequences on demand, as described in the FLASH project overview.
🔗 Read original →
The American agency ARPA-H has commissioned GE HealthCare to assemble a system for on-demand DNA production and verification. On August 4, DNA Script, a developer of enzyme-based DNA synthesis technology, announced its participation in the four-year FLASH project. ARPA-H has contracted with primary contractor GE HealthCare for up to $26 million; the agency's card indicates the project began on March 16.
The team is to assemble an automated system for DNA manufacturing, purification, and verification, which will be transferred to the program's early users. A researcher starts with a digital sequence, but for the experience, they need a physical DNA molecule. It needs to be manufactured, purified from byproducts, and verified to match the ordered sequence. According to ARPA-H's card, the FLASH project is described as an automated system that should connect these operations.
The first step in this path is DNA Script's enzyme-based synthesis. Enzymes are proteins that initiate chemical reactions; in the company's technology, one such enzyme adds one nucleotide, the DNA building block, to the growing chain at a time. A temporary blocking group on the added nucleotide stops the chain elongation. It is removed, and the enzyme adds the next link. The sequence is assembled one nucleotide per cycle. In the project, DNA Script is responsible for enzyme-based synthesis, while GE HealthCare adds scaling technology to increase production volume.
The program should connect synthesis with purification, verification of the finished DNA's match to the ordered sequence, and automation. In June, synthetic DNA suppliers were called to verify both the ordered sequence and the buyer. For FLASH, ARPA-H requires biosecurity and cybersecurity in addition to production stages. According to the agency's plan, after development, the system will be transferred to the program's early users: researchers working on DNA-based medicines will be able to produce the necessary sequences on demand, as described in the FLASH project overview.
🔗 Read original →
PubMed Central (PMC)
Sequence Preference and Initiator Promiscuity for De Novo DNA Synthesis by Terminal Deoxynucleotidyl Transferase
The untemplated activity of terminal deoxynucleotidyl transferase (TdT) represents its most appealing feature. Its use is well established in applications aiming for extension of a DNA initiator strand, but a more recent focus points to its ...
AI Lab Hub
Gladstone and Stanford University are creating an AI hub to model the flow of laboratory experience. On August 6, Le Cong and Kathy Pollard announced the creation of a joint AI hub between Gladstone and Stanford University. A related preprint, published on August 4, describes an AI system that stores all important information about the experiment in one constantly updated record.
The AI model can suggest increasing the concentration of a reagent, and the instrument can add it to the samples. However, such a step requires data that is often stored in the human mind and in different files: the laboratory technician replaced a batch of reagent, the instrument needs calibration, the protocol was changed, and the new result requires re-verification.
Without this information, the model builds a plan based on an incomplete description of the laboratory. The authors of the preprint call such a shared memory model a laboratory state model. It should store information about samples, reagents, instruments, protocols, observations, human interventions, and uncertainty.
According to this scheme, a robot receives a command to increase the concentration of a reagent only after the system checks the sample volume, available reagent batch, instrument calibration, and protocol limitations. Then it prepares the next step allowed by the protocol or passes the question to the scientist.
The new text adds to this video data from instruments, analysis results, and scientist decisions: all of them should update the common record of the experiment state. The authors propose testing such a system in a scenario that simulates the course of an experiment: can it link data from different tools, propose a testable hypothesis, reliably execute a plan, and timely pass the decision to a human.
The system should recognize the normal course of work, an erroneous signal in the data, a failure that can be corrected, or a moment when it is necessary to repeat the experiment, change the protocol, or stop the work. In this architecture, the scientist asks the question, interprets the result, and makes unexpected, ambiguous, and high-risk decisions, as described in the preprint published in August.
🔗 Read original →
Gladstone and Stanford University are creating an AI hub to model the flow of laboratory experience. On August 6, Le Cong and Kathy Pollard announced the creation of a joint AI hub between Gladstone and Stanford University. A related preprint, published on August 4, describes an AI system that stores all important information about the experiment in one constantly updated record.
The AI model can suggest increasing the concentration of a reagent, and the instrument can add it to the samples. However, such a step requires data that is often stored in the human mind and in different files: the laboratory technician replaced a batch of reagent, the instrument needs calibration, the protocol was changed, and the new result requires re-verification.
Without this information, the model builds a plan based on an incomplete description of the laboratory. The authors of the preprint call such a shared memory model a laboratory state model. It should store information about samples, reagents, instruments, protocols, observations, human interventions, and uncertainty.
According to this scheme, a robot receives a command to increase the concentration of a reagent only after the system checks the sample volume, available reagent batch, instrument calibration, and protocol limitations. Then it prepares the next step allowed by the protocol or passes the question to the scientist.
The new text adds to this video data from instruments, analysis results, and scientist decisions: all of them should update the common record of the experiment state. The authors propose testing such a system in a scenario that simulates the course of an experiment: can it link data from different tools, propose a testable hypothesis, reliably execute a plan, and timely pass the decision to a human.
The system should recognize the normal course of work, an erroneous signal in the data, a failure that can be corrected, or a moment when it is necessary to repeat the experiment, change the protocol, or stop the work. In this architecture, the scientist asks the question, interprets the result, and makes unexpected, ambiguous, and high-risk decisions, as described in the preprint published in August.
🔗 Read original →
X (formerly Twitter)
Le Cong@Stanford, AI+Bio+Gene-Editing (@lecong) on X
Thrilled to release two new preprints on intelligent labs for driving science and innovation. This is in close coordination with Aviv Regev, Jian Ma (@jmuiuc), Michelle Lee (@michellearning), and the teams at @Genentech, @SCSatCMU, and @Princeton University.…