DNA methylation blood study shows most CpG links stable with age, some reconfigured
On 18 September researchers posted a preprint on DNA methylation in whole blood from more than 7500 individuals aged 17 to 99. They focused on CpG sites and defined modules as groups of sites that usually change together.
After removing the average age‑related methylation shift, they built a module network from the youngest group (<36 years) and compared it to seven older groups, isolating changes in connections between sites rather than overall shifts.
The network contained 69 modules; they kept the 53 largest for further analysis. About 70% of these modules showed no significant change in connectivity with age, while 11 modules lost more than 5% of their connectivity from the youngest to the oldest group.
Stable modules tended to lie near gene promoters and in long‑range DNA contacts, whereas changing modules were found in late‑replicating, tightly packed chromatin. The approach builds on work from 2023 that compared similar methylation modules across 348 mammalian species and linked them to biological traits.
🔗 Read original →
On 18 September researchers posted a preprint on DNA methylation in whole blood from more than 7500 individuals aged 17 to 99. They focused on CpG sites and defined modules as groups of sites that usually change together.
After removing the average age‑related methylation shift, they built a module network from the youngest group (<36 years) and compared it to seven older groups, isolating changes in connections between sites rather than overall shifts.
The network contained 69 modules; they kept the 53 largest for further analysis. About 70% of these modules showed no significant change in connectivity with age, while 11 modules lost more than 5% of their connectivity from the youngest to the oldest group.
Stable modules tended to lie near gene promoters and in long‑range DNA contacts, whereas changing modules were found in late‑replicating, tightly packed chromatin. The approach builds on work from 2023 that compared similar methylation modules across 348 mammalian species and linked them to biological traits.
🔗 Read original →
CFTR modulator reverses biological aging in cystic fibrosis patients
Eight previously untreated cystic fibrosis patients received elexacaftor/tezacaftor/ivacaftor (ETI) for one year; their median epigenetic age acceleration shifted from +1.92 years to -1.70 years (p=0.008). The findings were reported in a preprint posted September 18, 2024 by researchers at the University of British Columbia.
ETI, marketed as Trikafta/Kaftrio, was approved in 2019 and is suitable for roughly 90% of CF patients; it markedly improves sweat chloride and lung function. In this cohort sweat chloride dropped from 98 to 41 mmol/L and lung function rose from 64.5% to 90% of normal.
Higher epigenetic age acceleration correlated with elevated blood IL‑6, IL‑1β and calprotectin, while DNAmGrimAge2—a clock trained to predict mortality and chronic disease—reflects lung aging. Improved lung function was linked to slower epigenetic aging, whereas higher sweat chloride pointed to faster aging.
An earlier Italian study of 52 adults showed a similar reduction in epigenetic age with ETI; this preprint replicates the observation using a different clock and directly ties the effect to inflammation markers. The synchronized changes in sweat chloride, lung function, inflammation and epigenetic age in the same eight individuals support the hypothesis that epigenetic aging in CF can reverse as rapidly as biochemical markers when CFTR function is restored.
🔗 Read original →
Eight previously untreated cystic fibrosis patients received elexacaftor/tezacaftor/ivacaftor (ETI) for one year; their median epigenetic age acceleration shifted from +1.92 years to -1.70 years (p=0.008). The findings were reported in a preprint posted September 18, 2024 by researchers at the University of British Columbia.
ETI, marketed as Trikafta/Kaftrio, was approved in 2019 and is suitable for roughly 90% of CF patients; it markedly improves sweat chloride and lung function. In this cohort sweat chloride dropped from 98 to 41 mmol/L and lung function rose from 64.5% to 90% of normal.
Higher epigenetic age acceleration correlated with elevated blood IL‑6, IL‑1β and calprotectin, while DNAmGrimAge2—a clock trained to predict mortality and chronic disease—reflects lung aging. Improved lung function was linked to slower epigenetic aging, whereas higher sweat chloride pointed to faster aging.
An earlier Italian study of 52 adults showed a similar reduction in epigenetic age with ETI; this preprint replicates the observation using a different clock and directly ties the effect to inflammation markers. The synchronized changes in sweat chloride, lung function, inflammation and epigenetic age in the same eight individuals support the hypothesis that epigenetic aging in CF can reverse as rapidly as biochemical markers when CFTR function is restored.
🔗 Read original →
PubMed Central (PMC)
Pulmonary Function Modulates Epigenetic Age in Subjects with Cystic Fibrosis
Cystic fibrosis (CF) is the most common severe autosomal recessive disease among Caucasians. Modulators of cystic fibrosis transmembrane conductance regulator (CFTR) mutated protein significantly improved the outcome of subjects with CF. In the ...
Mathematical control theory reframes aging as loss of safe manageability
The article by Alex Zhavoronkov, founder and CEO of Insilico Medicine, and mathematician Bad Mishra of the Courant Institute at New York University was published in Aging, September 16, 2026.
In their model the body is a point in the space of possible biological states, while drugs and other interventions are forces that push this point in different directions. Biological age is defined as the minimal cost in dose, risk, and time required to return the body to the zone of normal function.
Existing aging theories — López‑Otín’s hallmarks, Aubrey de Grey’s SENS, geroscience, and Blagosklonny’s hyper‑functional theory — describe what changes in cells and tissues with age, but they do not answer the practical question of which intervention, at what dose, and in what sequence to apply to a given organism now.
Six months earlier Zhavoronkov announced that aging clocks are dying and
🔗 Read original →
The article by Alex Zhavoronkov, founder and CEO of Insilico Medicine, and mathematician Bad Mishra of the Courant Institute at New York University was published in Aging, September 16, 2026.
In their model the body is a point in the space of possible biological states, while drugs and other interventions are forces that push this point in different directions. Biological age is defined as the minimal cost in dose, risk, and time required to return the body to the zone of normal function.
Existing aging theories — López‑Otín’s hallmarks, Aubrey de Grey’s SENS, geroscience, and Blagosklonny’s hyper‑functional theory — describe what changes in cells and tissues with age, but they do not answer the practical question of which intervention, at what dose, and in what sequence to apply to a given organism now.
Six months earlier Zhavoronkov announced that aging clocks are dying and
🔗 Read original →
Nature
Longitudinal analysis of blood markers reveals progressive loss of resilience and predicts human lifespan limit
Nature Communications - Aging is associated with an increased risk of chronic diseases and functional decline. Here, the authors investigate the fluctuations of physiological indices along aging...
UK Study Links Neighborhood Deprivation, Crime, and Pollution to Faster Biological Aging
A large British study posted as a preprint on medRxiv on 16 September examined how 30 aspects of the residential environment relate to biological aging in 3307 adults from the UK Understanding Society cohort. The work was led by Gergö Baranyi of University College London and Cathryn Tonne of the Barcelona Institute for Global Health. They used methylation‑based DNA clocks to measure biological age, focusing on white participants only.
Five age‑estimation tests and four analytical methods converged on the same finding: the newest clocks that measure the pace of aging — especially DunedinPACE — showed the strongest links to environmental stressors. Older clocks such as Horvath and Hannum responded weakly, while PhenoAge showed intermediate sensitivity. Notably, the
🔗 Read original →
A large British study posted as a preprint on medRxiv on 16 September examined how 30 aspects of the residential environment relate to biological aging in 3307 adults from the UK Understanding Society cohort. The work was led by Gergö Baranyi of University College London and Cathryn Tonne of the Barcelona Institute for Global Health. They used methylation‑based DNA clocks to measure biological age, focusing on white participants only.
Five age‑estimation tests and four analytical methods converged on the same finding: the newest clocks that measure the pace of aging — especially DunedinPACE — showed the strongest links to environmental stressors. Older clocks such as Horvath and Hannum responded weakly, while PhenoAge showed intermediate sensitivity. Notably, the
🔗 Read original →
medRxiv
Multidomain residential external exposome and epigenetic ageing in the UK general population
Residential environmental exposures might be associated with biological ageing, but evidence has largely focused on a limited set of single exposures. We investigated associations between the multidomain residential external exposome and epigenetic ageing…
LDDM generative model designs binders for five proteins, with X‑ray confirmation of one predicted pose
A preprint from the EPFL team describing LDDM — a generative model that creates drug‑like molecules fitted to a target protein structure — was released on bioRxiv, September 18. The authors synthesized and tested the model’s proposals for five different proteins, confirming binding in all five cases; for one target the X‑ray pose matched the prediction almost exactly.
LDDM was trained on 836 thousand examples of hidden molecular fragments, a dataset 16 times larger than the entire collection of deciphered protein‑ligand structures in the main global database. By limiting each growth step to chemically purchasable blocks, the model builds synthesizable molecules directly during generation.
In the molecular‑glue experiment linking CDO1 to VHL for degradation, 67% of the LDDM‑designed variants retained nanomolar binding affinity. For a peptide inhibitor of cathepsin S — a cancer‑related enzyme on which the model had never been trained — a single‑amino‑acid substitution improved binding 3.5‑fold, from 70.4 nM to 19.9 nM.
For the metabolic enzyme PGK1 (linked to cancer, Parkinson’s and ALS), the best LDDM molecule bound at 14.7 µM, stronger than the control compound at 30 µM as verified by NMR. LDDM also produced a BRD4 binder with an affinity of 40–60 µM, whose NMR‑observed binding matched the model’s prediction. In the SARS‑CoV‑2 Mac1 challenge, LDDM screened only 57 molecules and identified a chemically distinct hit at 46.2 µM; X‑ray analysis confirmed its pose with an accuracy of 1.4–2.3 Å, whereas a public competition of 23 teams had yielded a best result near 18 µM after testing 1739 compounds.
The model failed to give a clear signal for KRAS, and for Pin1 the crystal pose differed from the prediction. Most hits remain below the potency needed for a drug candidate, which the authors view as a pragmatic compromise enabling experimental testing of designed molecules. Lead author Ilya Igashov — an MIPT graduate and EPFL PhD student under Bruno Correia and Oxford professor Michael Bronstein — notes that LDDM continues the lab’s four‑year trajectory from fragment‑based assembly to de‑novo design and synthesizability prediction.
🔗 Read original →
A preprint from the EPFL team describing LDDM — a generative model that creates drug‑like molecules fitted to a target protein structure — was released on bioRxiv, September 18. The authors synthesized and tested the model’s proposals for five different proteins, confirming binding in all five cases; for one target the X‑ray pose matched the prediction almost exactly.
LDDM was trained on 836 thousand examples of hidden molecular fragments, a dataset 16 times larger than the entire collection of deciphered protein‑ligand structures in the main global database. By limiting each growth step to chemically purchasable blocks, the model builds synthesizable molecules directly during generation.
In the molecular‑glue experiment linking CDO1 to VHL for degradation, 67% of the LDDM‑designed variants retained nanomolar binding affinity. For a peptide inhibitor of cathepsin S — a cancer‑related enzyme on which the model had never been trained — a single‑amino‑acid substitution improved binding 3.5‑fold, from 70.4 nM to 19.9 nM.
For the metabolic enzyme PGK1 (linked to cancer, Parkinson’s and ALS), the best LDDM molecule bound at 14.7 µM, stronger than the control compound at 30 µM as verified by NMR. LDDM also produced a BRD4 binder with an affinity of 40–60 µM, whose NMR‑observed binding matched the model’s prediction. In the SARS‑CoV‑2 Mac1 challenge, LDDM screened only 57 molecules and identified a chemically distinct hit at 46.2 µM; X‑ray analysis confirmed its pose with an accuracy of 1.4–2.3 Å, whereas a public competition of 23 teams had yielded a best result near 18 µM after testing 1739 compounds.
The model failed to give a clear signal for KRAS, and for Pin1 the crystal pose differed from the prediction. Most hits remain below the potency needed for a drug candidate, which the authors view as a pragmatic compromise enabling experimental testing of designed molecules. Lead author Ilya Igashov — an MIPT graduate and EPFL PhD student under Bruno Correia and Oxford professor Michael Bronstein — notes that LDDM continues the lab’s four‑year trajectory from fragment‑based assembly to de‑novo design and synthesizability prediction.
🔗 Read original →
Nature
Equivariant 3D-conditional diffusion model for molecular linker design
Nature Machine Intelligence - Fragment-based molecular design uses chemical motifs and combines them into bio-active compounds. While this approach has grown in capability, molecular linker methods...
Bone Healing Model Shows Why Osteoporosis Drugs Lose Effect Over Time
Reviews in Endocrine and Metabolic Disorders published a review on 17 September 2026 from the University of California, Davis lab. The authors describe bone recovery as a coordinated action of hormones, immune cells, and the local stem‑cell niche, and show that aging disrupts this coordination at two distinct, partially reversible levels.
A precise definition of skeletal stem cells emerged in 2018 when a Stanford team isolated and functionally validated such cells in humans using surface markers; that work has been cited over 600 times. Before this, the field relied on the ill‑defined term “mesenchymal stem cells,” a heterogeneous mix lacking reliable links to outcomes.
One of the 2018 discoverers, Thomas Ambrosi, moved to Davis in 2024 and, with co‑author Kun Chen in 2025, demonstrated that human skeletal stem cells exist in at least four states; the state that activates depends on the cell’s anatomical niche and its specific microenvironment.
Bone itself secretes hormones such as osteocalcin and FGF23 that influence brain, kidney, and metabolism. Fracture healing proceeds through an active hematoma: neutrophils build a temporary matrix within 48 hours, then macrophages pass through four states from inflammatory to reparative. Administering mice the CSF1R blocker pexidartinib from days 3 to 14 after fracture reduces macrophage accumulation, lessens scarring, and accelerates union.
Aging affects the system on two levels. Young blood rejuvenates many tissues—parabiosis of old and young mice extends old‑mouse lifespan by 10%—but does not restore skeletal stem‑cell function; Ambrosi’s earlier work showed these cells lose bone‑forming capacity and are not rescued by young‑blood infusion or hematopoietic transplant. In contrast, the niche defect is reversible: declining tenascin C with age impairs macrophage recruitment to injury, and restoring this protein in mice renews old animals’ bone‑healing ability.
Osteoporosis afflicts one third of women and one fifth of men over 50 according to the International Osteoporosis Foundation, leading to as many as up to 37 million fractures per year. Estrogen deficiency directly harms skeletal stem cells, while glucocorticoids disrupt bone‑vascular coupling via the protein basigin; blocking basigin with an antibody prevents bone loss in mice (demonstrated by
🔗 Read original →
Reviews in Endocrine and Metabolic Disorders published a review on 17 September 2026 from the University of California, Davis lab. The authors describe bone recovery as a coordinated action of hormones, immune cells, and the local stem‑cell niche, and show that aging disrupts this coordination at two distinct, partially reversible levels.
A precise definition of skeletal stem cells emerged in 2018 when a Stanford team isolated and functionally validated such cells in humans using surface markers; that work has been cited over 600 times. Before this, the field relied on the ill‑defined term “mesenchymal stem cells,” a heterogeneous mix lacking reliable links to outcomes.
One of the 2018 discoverers, Thomas Ambrosi, moved to Davis in 2024 and, with co‑author Kun Chen in 2025, demonstrated that human skeletal stem cells exist in at least four states; the state that activates depends on the cell’s anatomical niche and its specific microenvironment.
Bone itself secretes hormones such as osteocalcin and FGF23 that influence brain, kidney, and metabolism. Fracture healing proceeds through an active hematoma: neutrophils build a temporary matrix within 48 hours, then macrophages pass through four states from inflammatory to reparative. Administering mice the CSF1R blocker pexidartinib from days 3 to 14 after fracture reduces macrophage accumulation, lessens scarring, and accelerates union.
Aging affects the system on two levels. Young blood rejuvenates many tissues—parabiosis of old and young mice extends old‑mouse lifespan by 10%—but does not restore skeletal stem‑cell function; Ambrosi’s earlier work showed these cells lose bone‑forming capacity and are not rescued by young‑blood infusion or hematopoietic transplant. In contrast, the niche defect is reversible: declining tenascin C with age impairs macrophage recruitment to injury, and restoring this protein in mice renews old animals’ bone‑healing ability.
Osteoporosis afflicts one third of women and one fifth of men over 50 according to the International Osteoporosis Foundation, leading to as many as up to 37 million fractures per year. Estrogen deficiency directly harms skeletal stem cells, while glucocorticoids disrupt bone‑vascular coupling via the protein basigin; blocking basigin with an antibody prevents bone loss in mice (demonstrated by
🔗 Read original →
PubMed Central (PMC)
An Interview with Cell Therapy Pioneer, Arnold Caplan
Diffusion model converts MRI to PET-like brain map, boosting dementia diagnosis accuracy
On September 17,
🔗 Read original →
On September 17,
🔗 Read original →
Cholesterol Switch Controls Lysosomal Mitochondrial Cleanup
Scientists identified a cholesterol‑dependent switch that determines whether lysosomes can fully digest damaged mitochondria. The finding was reported Nature Communications, 16 September by the laboratory of Yasuori Saheki at Nanyang Technological University, Singapore.
The lysosome normally contains little cholesterol; the enzyme PI4KIIα marks its membrane with a lipid tag that recruits the transporter OSBP to pump cholesterol from the endoplasmic reticulum. This influx is required to maintain lysosomal acidity, membrane integrity, and to complete the digestion process.
The cholesterol supplied comes from the ER, not from the damaged mitochondrion itself, and its transport activates SREBP-2, which synthesizes new cholesterol to replenish both the lysosome and ER stores. Earlier work from the same lab showed that OSBP hyperactivation lowers ER cholesterol and triggers SREBP-2 activation.
When cells were deprived of cholesterol, lysosomes digested damaged mitochondria poorly, lost acidity, and their membranes became more fragile; restoring cholesterol rescued all three defects, demonstrating that cholesterol influx is essential for complete digestion.
Free fatty acids released from the broken mitochondria are converted by DGAT1 into neutral fat and stored in lipid droplets, preventing cytoplasmic toxicity; this pathway is activated specifically during mitochondrial damage (e.g., iron depletion) but not during general starvation.
The authors link the mechanism to neurodegeneration, noting that mutations in PINK1 and Parkin cause hereditary Parkinson’s disease, and lysosomal dysfunction is observed in Alzheimer’s where lysosomal pathway risk scores match brain damage; the connection is drawn from independent literature on lysosomal defects in these diseases.
Each component of the pathway — PI4KIIα, OSBP, and SREBP-2 — was sequentially inhibited and then cholesterol was re‑added; the digestive defect was rescued each time, confirming a causal role for every step. Experiments were performed in skin cell cultures and other cell lines.
🔗 Read original →
Scientists identified a cholesterol‑dependent switch that determines whether lysosomes can fully digest damaged mitochondria. The finding was reported Nature Communications, 16 September by the laboratory of Yasuori Saheki at Nanyang Technological University, Singapore.
The lysosome normally contains little cholesterol; the enzyme PI4KIIα marks its membrane with a lipid tag that recruits the transporter OSBP to pump cholesterol from the endoplasmic reticulum. This influx is required to maintain lysosomal acidity, membrane integrity, and to complete the digestion process.
The cholesterol supplied comes from the ER, not from the damaged mitochondrion itself, and its transport activates SREBP-2, which synthesizes new cholesterol to replenish both the lysosome and ER stores. Earlier work from the same lab showed that OSBP hyperactivation lowers ER cholesterol and triggers SREBP-2 activation.
When cells were deprived of cholesterol, lysosomes digested damaged mitochondria poorly, lost acidity, and their membranes became more fragile; restoring cholesterol rescued all three defects, demonstrating that cholesterol influx is essential for complete digestion.
Free fatty acids released from the broken mitochondria are converted by DGAT1 into neutral fat and stored in lipid droplets, preventing cytoplasmic toxicity; this pathway is activated specifically during mitochondrial damage (e.g., iron depletion) but not during general starvation.
The authors link the mechanism to neurodegeneration, noting that mutations in PINK1 and Parkin cause hereditary Parkinson’s disease, and lysosomal dysfunction is observed in Alzheimer’s where lysosomal pathway risk scores match brain damage; the connection is drawn from independent literature on lysosomal defects in these diseases.
Each component of the pathway — PI4KIIα, OSBP, and SREBP-2 — was sequentially inhibited and then cholesterol was re‑added; the digestive defect was rescued each time, confirming a causal role for every step. Experiments were performed in skin cell cultures and other cell lines.
🔗 Read original →
Nature
Cholesterol maintains the degradative capacity of lysosomes during clearance and recycling of dysfunctional mitochondria
Nature Communications - Damaged mitochondria are cleared by lysosomes. Here, authors show that ER-to-lysosome cholesterol transport preserves lysosomal degradative function during mitophagy,...
AI Agent Evaluates Severity of All Human Genetic Disease Traits
An autonomous AI agent built on GPT‑4o independently assessed the severity of all 10,211 Human Phenotype Ontology terms by scanning PubMed and applying ACMG criteria, with the work released as a preprint on Sep 17, 2025 preprint, Sep 17, 2025. The agent aggregated these evaluations to the level of 8,738 gene‑disease pairs.
To ensure verifiability, the team used a ReAct framework: the agent must cite a source for every conclusion and cannot rely on internal model knowledge. After testing six models, including GPT‑5 and Grok‑4, they selected GPT‑4o for its 92% accuracy and balanced reasoning. A separate controller agent weights evidence and forwarded 1,161 low‑confidence assessments to human reviewers.
On a set of 941 phenotypes independently labeled by two geneticists, the agent agreed in 93.55% of cases. At the gene‑disease level, it classified 79.5% of the 8,738 pairs as severe or very severe. The resulting severe‑gene list matched the Australian Mackenzie’s Mission carrier‑screening panel at 95.2% and the ACMG carrier panel at 99.3% New England Journal of Medicine, 2024.
Processing one trait costs about 10 cents, so the full set of 10,211 terms required roughly $1,000—a fraction of the years needed by expert committees. The agent reproduced independent severity judgments for 29 genes in 20 cases, assigned a stricter category in eight, and under‑rated only one gene (BCKDHB). About a third of terms (e.g., headache) could not be placed on the severity scale. Rights to the technology belong to UNSW Sydney and are partially licensed to 23Strands.
🔗 Read original →
An autonomous AI agent built on GPT‑4o independently assessed the severity of all 10,211 Human Phenotype Ontology terms by scanning PubMed and applying ACMG criteria, with the work released as a preprint on Sep 17, 2025 preprint, Sep 17, 2025. The agent aggregated these evaluations to the level of 8,738 gene‑disease pairs.
To ensure verifiability, the team used a ReAct framework: the agent must cite a source for every conclusion and cannot rely on internal model knowledge. After testing six models, including GPT‑5 and Grok‑4, they selected GPT‑4o for its 92% accuracy and balanced reasoning. A separate controller agent weights evidence and forwarded 1,161 low‑confidence assessments to human reviewers.
On a set of 941 phenotypes independently labeled by two geneticists, the agent agreed in 93.55% of cases. At the gene‑disease level, it classified 79.5% of the 8,738 pairs as severe or very severe. The resulting severe‑gene list matched the Australian Mackenzie’s Mission carrier‑screening panel at 95.2% and the ACMG carrier panel at 99.3% New England Journal of Medicine, 2024.
Processing one trait costs about 10 cents, so the full set of 10,211 terms required roughly $1,000—a fraction of the years needed by expert committees. The agent reproduced independent severity judgments for 29 genes in 20 cases, assigned a stricter category in eight, and under‑rated only one gene (BCKDHB). About a third of terms (e.g., headache) could not be placed on the severity scale. Rights to the technology belong to UNSW Sydney and are partially licensed to 23Strands.
🔗 Read original →
PubMed Central (PMC)
The promises and pitfalls of automated variant interpretation: a comprehensive review
The interpretation of DNA variants enables personalized medicine through precise diagnosis and treatment selection. To address the challenges of manual interpretation, a wide range of automated tools has been created. This study evaluates these ...
AlphaGenome and AlphaMissense assess somatic mutation impact across 8800 tumors
On 16 September, the Ilias Georgakopoulos‑Soares lab at the University of Texas at Austin posted a preprint on medRxiv evaluating AlphaGenome and AlphaMissense. The models were applied to 8800 patients with 33 cancer types from The Cancer Genome Atlas (TCGA) to measure gene damage from somatic mutations. In patients lacking known “hotspot” driver mutations, this damage correlated with survival and treatment outcome in an independent cohort of 570 tumors.
Oncogenomics
🔗 Read original →
On 16 September, the Ilias Georgakopoulos‑Soares lab at the University of Texas at Austin posted a preprint on medRxiv evaluating AlphaGenome and AlphaMissense. The models were applied to 8800 patients with 33 cancer types from The Cancer Genome Atlas (TCGA) to measure gene damage from somatic mutations. In patients lacking known “hotspot” driver mutations, this damage correlated with survival and treatment outcome in an independent cohort of 570 tumors.
Oncogenomics
🔗 Read original →
medRxiv
Genomic foundation model-derived disruption profiling links somatic mutations to cancer biology and clinical outcomes
Cancer genomics has concentrated on individual mutations, overlooking whether somatic mutations can accumulate to produce partial, gene-level disruption with biological and clinical consequences. Sequence-to-function models can quantify these effects directly…
Peptide C43-4 Activates APC/C-Cdh1 to Extend Lifespan in Yeast and Worms
The peptide C43-4 was identified as the first activator of the APC/C-Cdh1 complex, a molecular machine that tags unnecessary proteins for destruction and thereby keeps cells in a healthy quiescent state. In yeast and C. elegans, C43-4 extended lifespan through a conserved longevity pathway involving the daf-16 (FOXO relative) and aak-2 (AMPK relative) genes.
APC functions by marking excess proteins for degradation; when paired with Cdc20 it drives cell division, whereas with Cdh1 it maintains a non‑dividing but viable quiescent state. The Cdh1 form deteriorates with age, causing APC substrates to accumulate—a change regarded as a molecular hallmark of aging, which the authors demonstrated in aging yeast cells.
Results were published 16 September in the journal Genetics. Chemical activators of the APC/Cdc20 branch are already tested as anticancer agents (without success); they prolonged the life of dividing yeast but not quiescent cells, which require a Cdh1‑specific activator.
The peptide library that yielded C43-4 originated from earlier work on reversing breast‑cancer drug resistance, where C43-4 also inhibited tumor growth in mice. Only C43-4 prolonged yeast lifespan in quiescence, and it did so exclusively via Cdh1. C43-4 contains a D‑box motif RxxL, the same signal used by natural APC substrates for recognition by Apc10.
Mutational analysis revealed that a single amino acid, D7 (located outside the D‑box), is essential; without D7 the peptide fails to bind Apc10 and loses its lifespan‑extending activity, indicating precise target recognition rather than nonspecific effects.
Synthetic
🔗 Read original →
The peptide C43-4 was identified as the first activator of the APC/C-Cdh1 complex, a molecular machine that tags unnecessary proteins for destruction and thereby keeps cells in a healthy quiescent state. In yeast and C. elegans, C43-4 extended lifespan through a conserved longevity pathway involving the daf-16 (FOXO relative) and aak-2 (AMPK relative) genes.
APC functions by marking excess proteins for degradation; when paired with Cdc20 it drives cell division, whereas with Cdh1 it maintains a non‑dividing but viable quiescent state. The Cdh1 form deteriorates with age, causing APC substrates to accumulate—a change regarded as a molecular hallmark of aging, which the authors demonstrated in aging yeast cells.
Results were published 16 September in the journal Genetics. Chemical activators of the APC/Cdc20 branch are already tested as anticancer agents (without success); they prolonged the life of dividing yeast but not quiescent cells, which require a Cdh1‑specific activator.
The peptide library that yielded C43-4 originated from earlier work on reversing breast‑cancer drug resistance, where C43-4 also inhibited tumor growth in mice. Only C43-4 prolonged yeast lifespan in quiescence, and it did so exclusively via Cdh1. C43-4 contains a D‑box motif RxxL, the same signal used by natural APC substrates for recognition by Apc10.
Mutational analysis revealed that a single amino acid, D7 (located outside the D‑box), is essential; without D7 the peptide fails to bind Apc10 and loses its lifespan‑extending activity, indicating precise target recognition rather than nonspecific effects.
Synthetic
🔗 Read original →
OUP Academic
Discovery of a small peptide that increases yeast lifespan by enhancing the function of APCCdh1 in nondividing quiescent cells
Abstract. Aging is accompanied by molecular hallmarks conserved from yeast to humans. One such hallmark, senescence, is an irreversible nondividing state t
Gene SIX2 drives fast muscle repair in a subset of stem cells
Different muscles in the body heal at different speeds after injury, and French and Hong Kong biologists have identified the gene SIX2 as the switch that puts a fraction of muscle stem cells into a rapid‑repair mode. In mice, a subset of skeletal‑muscle stem cells expresses the embryonic gene
🔗 Read original →
Different muscles in the body heal at different speeds after injury, and French and Hong Kong biologists have identified the gene SIX2 as the switch that puts a fraction of muscle stem cells into a rapid‑repair mode. In mice, a subset of skeletal‑muscle stem cells expresses the embryonic gene
🔗 Read original →
Dyskerin self‑pairing uncovered as key defect in telomerase RNA binding
Dyskerin forms a homodimer that does not require telomerase RNA, and the four X‑linked mutations that cause congenital dyskeratosis leave this dimer intact while weakening the protein’s grip on the telomerase RNA component (hTR). The finding was reported by a team from McGill University on 16 September in the journal RNA.
Each cell division shortens telomeres, and when they become critically short the cell stops dividing – the limit described by Leonard Hayflick in 1961. Telomerase can rebuild telomeres, and the elucidation of this mechanism earned Elizabeth Blackburn, Carol Greider and Jack Szostak the Nobel Prize in Physiology or Medicine in 2009. For telomerase to work, the cell must assemble its RNA subunit hTR, a task performed by dyskerin; without dyskerin hTR is degraded before it can function.
Structural images had previously shown two dyskerin molecules touching at a site
🔗 Read original →
Dyskerin forms a homodimer that does not require telomerase RNA, and the four X‑linked mutations that cause congenital dyskeratosis leave this dimer intact while weakening the protein’s grip on the telomerase RNA component (hTR). The finding was reported by a team from McGill University on 16 September in the journal RNA.
Each cell division shortens telomeres, and when they become critically short the cell stops dividing – the limit described by Leonard Hayflick in 1961. Telomerase can rebuild telomeres, and the elucidation of this mechanism earned Elizabeth Blackburn, Carol Greider and Jack Szostak the Nobel Prize in Physiology or Medicine in 2009. For telomerase to work, the cell must assemble its RNA subunit hTR, a task performed by dyskerin; without dyskerin hTR is degraded before it can function.
Structural images had previously shown two dyskerin molecules touching at a site
🔗 Read original →
rnajournal.cshlp.org
Disease mutations in the N-terminal extension and α2′-helix of dyskerin do not abolish dimerization but disrupt binding to
…
…
A monthly journal publishing high-quality, peer-reviewed research on all topics related to RNA and its metabolism in all organisms
New MRI‑visible brain aging marker identified: choroid plexus calcification linked to metabolic risks
A preprint from 16 September analysed 30,012 MRI scans from the UK Biobank, measuring choroid plexus calcification with quantitative susceptibility mapping (QSM). Calcification rose with age and was higher in men than women.
The strongest association was found in the third ventricle, where calcification correlated most tightly with diabetes, obesity, smoking, hypertension and ischemic heart disease, even after adjusting for ventricular volume. This pattern was weaker in the
🔗 Read original →
A preprint from 16 September analysed 30,012 MRI scans from the UK Biobank, measuring choroid plexus calcification with quantitative susceptibility mapping (QSM). Calcification rose with age and was higher in men than women.
The strongest association was found in the third ventricle, where calcification correlated most tightly with diabetes, obesity, smoking, hypertension and ischemic heart disease, even after adjusting for ventricular volume. This pattern was weaker in the
🔗 Read original →
PubMed Central (PMC)
Age-related Vascular Alterations in the Choroid Plexus: Novel Insights from Pathophysiology and Imaging Studies
The choroid plexus (ChP), a highly vascularized brain structure responsible for cerebrospinal fluid (CSF) production, undergoes significant age-related changes that may contribute to neurodegenerative diseases involving disrupted immune regulation, ...
MIT discovers common lipid aging signal in mouse lung and skin
On 21 September 2026, scientists from MIT and Massachusetts General Hospital published in Nature Aging a method called RamanOmics that combines label‑free Raman microscopy with RNA sequencing of the same tissue sections. The approach lets them study chemical signals without destroying cells.
In mouse lung and skin they found different sets of active genes with age, but a single chemical signal of aging appeared in both tissues. With age some cells enter senescence, stop dividing, linger, and affect neighbours through inflammation and tissue remodeling.
Senescent cells are usually identified by markers p21 and p16, yet neither is definitive because each appears in various cell types and stages. The authors chose p21 as the primary marker because removing p21‑positive cells prevents radiation‑induced bone loss in mice.
Earlier, the NIH suggested describing such cells by a set of coordinates that include chemical makeup; RamanOmics now provides that measurement for the first time. In 2024 MIT showed that a Raman image of a living cell can predict its transcriptional profile.
Researchers led by Jae Zhang, Salvatore Sorrentino and colleagues linked that prediction to RNA‑seq and the spatial gene‑activity map STARmap on lung and skin slices from 2‑month‑old and 26‑month‑old mice. Old mouse lungs showed inflammation and repair genes; skin showed keratinization genes — the gene sets diverged.
Yet in both tissues the same Raman peak at 1131–1135 cm⁻¹ intensified, indicating a specific lipid class that serves as the first common aging signature despite different genetic “signatures” of the organs.
From the most informative Raman peaks and genes the team built a “barcode” and trained a classifier to distinguish senescent from normal cells more accurately than any single signal. Accuracy rose from 73.7% to 77.6% in lung and from 61.8% to 65.5% in skin.
A test on a mouse skin wound excluded chance: three days after injury p21 activity increased together with the same keratinization genes and lipid peak seen during aging.
Salvatore Sorrentino said, “By combining the most important Raman features with the most important genetic signatures we created a barcode that helps find aging cells more objectively.” Jae Zhang added that one could envision an endoscope that looks inside the body and detects cellular senescence.
For now the method works only on mice; analysing about 1 mm² of tissue takes roughly 30 hours. Senescent cells are rare — fractions of a percent — so the finding is still correlative. Causality will be tested by disabling the enzymes that lengthen the lipids.
Because the aging signal is visible in scattered light, not in RNA (which requires destroying the cell), the approach can be sped up and transferred to living tissue, paving the way for non‑destructive monitoring of senescent‑cell burden and for testing whether senolytics lower that burden.
🔗 Read original →
On 21 September 2026, scientists from MIT and Massachusetts General Hospital published in Nature Aging a method called RamanOmics that combines label‑free Raman microscopy with RNA sequencing of the same tissue sections. The approach lets them study chemical signals without destroying cells.
In mouse lung and skin they found different sets of active genes with age, but a single chemical signal of aging appeared in both tissues. With age some cells enter senescence, stop dividing, linger, and affect neighbours through inflammation and tissue remodeling.
Senescent cells are usually identified by markers p21 and p16, yet neither is definitive because each appears in various cell types and stages. The authors chose p21 as the primary marker because removing p21‑positive cells prevents radiation‑induced bone loss in mice.
Earlier, the NIH suggested describing such cells by a set of coordinates that include chemical makeup; RamanOmics now provides that measurement for the first time. In 2024 MIT showed that a Raman image of a living cell can predict its transcriptional profile.
Researchers led by Jae Zhang, Salvatore Sorrentino and colleagues linked that prediction to RNA‑seq and the spatial gene‑activity map STARmap on lung and skin slices from 2‑month‑old and 26‑month‑old mice. Old mouse lungs showed inflammation and repair genes; skin showed keratinization genes — the gene sets diverged.
Yet in both tissues the same Raman peak at 1131–1135 cm⁻¹ intensified, indicating a specific lipid class that serves as the first common aging signature despite different genetic “signatures” of the organs.
From the most informative Raman peaks and genes the team built a “barcode” and trained a classifier to distinguish senescent from normal cells more accurately than any single signal. Accuracy rose from 73.7% to 77.6% in lung and from 61.8% to 65.5% in skin.
A test on a mouse skin wound excluded chance: three days after injury p21 activity increased together with the same keratinization genes and lipid peak seen during aging.
Salvatore Sorrentino said, “By combining the most important Raman features with the most important genetic signatures we created a barcode that helps find aging cells more objectively.” Jae Zhang added that one could envision an endoscope that looks inside the body and detects cellular senescence.
For now the method works only on mice; analysing about 1 mm² of tissue takes roughly 30 hours. Senescent cells are rare — fractions of a percent — so the finding is still correlative. Causality will be tested by disabling the enzymes that lengthen the lipids.
Because the aging signal is visible in scattered light, not in RNA (which requires destroying the cell), the approach can be sped up and transferred to living tissue, paving the way for non‑destructive monitoring of senescent‑cell burden and for testing whether senolytics lower that burden.
🔗 Read original →
Nature
RamanOmics decodes the spatial vibrational–molecular architecture of senescence in aging and repair
Nature Aging - Zhang, Chen, Monticolo, Sorrentino and colleagues pair label-free Raman imaging with spatial and single-cell transcriptomics to probe biochemical signatures of senescence in aging...
Menopausal Hormone Therapy Effects on Brain Depend on APOEε4 Gene and Estrogen Formulation
A large analysis of the NACC database, which includes 8741 women from over 40 U.S. dementia centers, found that the link between menopausal hormone therapy and memory or brain structure depends on both the APOEε4 gene and the estrogen formulation used.
In 2003 the Women's Health Initiative reported that women 65 years and older taking conjugated equine estrogens with medroxyprogesterone (Premarin/Prempro) had nearly double the risk of probable dementia, a risk ratio 2.05; later observational studies of estradiol suggested the opposite benefit and hinted at a role for APOEε4.
A preprint released on 20 September from Liisa Galea’s lab at CAMH in Toronto directly compared formulations within one cohort, dividing participants (average age ≈70) into five groups: no therapy, estradiol alone, estradiol + progestogen, conjugated equine estrogens (CEE) alone, and CEE + medroxyprogesterone.
After statistical adjustment for age and risk factors, estradiol showed stronger benefits for APOEε4 carriers, whereas CEE appeared more helpful for non‑carriers; CEE was linked to poorer episodic memory than estradiol or no therapy but to better verbal fluency.
The authors note that CEE mainly contains a weak estrogen with low brain‑receptor affinity, while estradiol binds strongly and has long been known to protect the hippocampus; the estradiol advantage was especially pronounced among Black participants despite their earlier, more severe menopause and lower therapy use.
In an MRI subsample of 930 women (871 untreated, 59 on estradiol), APOEε4 carriers receiving estradiol had thicker cortex in four AD‑vulnerable regions, a larger hippocampus, and fewer white‑matter lesion spots than untreated carriers; no such pattern was seen in non‑carriers.
The researchers conclude that these results reframe decades of contradictory hormone‑
🔗 Read original →
A large analysis of the NACC database, which includes 8741 women from over 40 U.S. dementia centers, found that the link between menopausal hormone therapy and memory or brain structure depends on both the APOEε4 gene and the estrogen formulation used.
In 2003 the Women's Health Initiative reported that women 65 years and older taking conjugated equine estrogens with medroxyprogesterone (Premarin/Prempro) had nearly double the risk of probable dementia, a risk ratio 2.05; later observational studies of estradiol suggested the opposite benefit and hinted at a role for APOEε4.
A preprint released on 20 September from Liisa Galea’s lab at CAMH in Toronto directly compared formulations within one cohort, dividing participants (average age ≈70) into five groups: no therapy, estradiol alone, estradiol + progestogen, conjugated equine estrogens (CEE) alone, and CEE + medroxyprogesterone.
After statistical adjustment for age and risk factors, estradiol showed stronger benefits for APOEε4 carriers, whereas CEE appeared more helpful for non‑carriers; CEE was linked to poorer episodic memory than estradiol or no therapy but to better verbal fluency.
The authors note that CEE mainly contains a weak estrogen with low brain‑receptor affinity, while estradiol binds strongly and has long been known to protect the hippocampus; the estradiol advantage was especially pronounced among Black participants despite their earlier, more severe menopause and lower therapy use.
In an MRI subsample of 930 women (871 untreated, 59 on estradiol), APOEε4 carriers receiving estradiol had thicker cortex in four AD‑vulnerable regions, a larger hippocampus, and fewer white‑matter lesion spots than untreated carriers; no such pattern was seen in non‑carriers.
The researchers conclude that these results reframe decades of contradictory hormone‑
🔗 Read original →
AI screens drug records, finds varenicline may cut dementia risk by half
Researchers at the VA Medical Center in Washington and George Washington University trained an AI model on about one million patient records without a prior hypothesis to find drugs that lower dementia risk. Varenicline, a smoking‑cessation drug approved in 2006, emerged as a strong candidate.
The team used a transformer‑type architecture to distinguish patients who later developed dementia from those who did not based on prescriptions, diagnoses, and lab results; interpretable AI then linked lower dementia risk to specific drug exposures. The findings were first reported in a preprint, 20 September.
In an independent validation cohort of 9,584 adults over 50 who were dementia‑free at baseline and started either varenicline
🔗 Read original →
Researchers at the VA Medical Center in Washington and George Washington University trained an AI model on about one million patient records without a prior hypothesis to find drugs that lower dementia risk. Varenicline, a smoking‑cessation drug approved in 2006, emerged as a strong candidate.
The team used a transformer‑type architecture to distinguish patients who later developed dementia from those who did not based on prescriptions, diagnoses, and lab results; interpretable AI then linked lower dementia risk to specific drug exposures. The findings were first reported in a preprint, 20 September.
In an independent validation cohort of 9,584 adults over 50 who were dementia‑free at baseline and started either varenicline
🔗 Read original →
medRxiv
Varenicline as a Repurposable Drug Candidate for ADRD Prevention
Objective: Effective strategies to prevent Alzheimer's disease and related dementias (ADRD) are limited. Implementing our Medication-Wide Association Study (MWAS+), we highlighted varenicline as a candidate. This study evaluated the association between varenicline…
Two Distinct Brain Pathways Protect Centenarians From Alzheimer's
Researchers at Amsterdam University Medical Center examined brain tissue from 112 centenarians who retained clear cognition into extreme age. They released their findings as a preprint, September 21.
The analysis showed that 44% of the donors had hardly accumulated amyloid‑beta, while 56% displayed amyloid levels typical of Alzheimer’s disease.
Among those with high amyloid, about 28% of the total cohort had tau spread into the presubiculum, entorhinal cortex and fusiform gyrus, mirroring the Alzheimer’s pattern and showing poorer memory. A similar proportion had comparable amyloid loads but tau barely entered those regions, and their cognition remained as sharp as in amyloid‑free individuals.
In a more detailed breakdown of 107 donors, 39% did not fit either the classic Alzheimer’s or the age‑related tau pattern, and in some high‑amyloid cases tau behaved like benign ageing. Cognition was significantly better for those with a stable, age‑like tau pattern than for those with an Alzheimer’s‑like or mixed pattern (p=0.004). The HLA‑II Hap‑B variant was linked to lower tau independent of amyloid load.
These results suggest two protective routes: preventing amyloid build‑up, or blocking its ability to trigger tau in the vulnerable medial‑temporal zones. A similar case was noted in a 115‑year‑old Dutch woman whose brain showed little amyloid or dementia‑related damage despite some tau presence.
🔗 Read original →
Researchers at Amsterdam University Medical Center examined brain tissue from 112 centenarians who retained clear cognition into extreme age. They released their findings as a preprint, September 21.
The analysis showed that 44% of the donors had hardly accumulated amyloid‑beta, while 56% displayed amyloid levels typical of Alzheimer’s disease.
Among those with high amyloid, about 28% of the total cohort had tau spread into the presubiculum, entorhinal cortex and fusiform gyrus, mirroring the Alzheimer’s pattern and showing poorer memory. A similar proportion had comparable amyloid loads but tau barely entered those regions, and their cognition remained as sharp as in amyloid‑free individuals.
In a more detailed breakdown of 107 donors, 39% did not fit either the classic Alzheimer’s or the age‑related tau pattern, and in some high‑amyloid cases tau behaved like benign ageing. Cognition was significantly better for those with a stable, age‑like tau pattern than for those with an Alzheimer’s‑like or mixed pattern (p=0.004). The HLA‑II Hap‑B variant was linked to lower tau independent of amyloid load.
These results suggest two protective routes: preventing amyloid build‑up, or blocking its ability to trigger tau in the vulnerable medial‑temporal zones. A similar case was noted in a 115‑year‑old Dutch woman whose brain showed little amyloid or dementia‑related damage despite some tau presence.
🔗 Read original →
medRxiv
Centenarians maintain cognition by resisting amyloid-β or decoupling it from tau propagation
Centenarians who maintain cognitive health provide a unique opportunity to investigate naturally occurring mechanisms that protect against Alzheimer's disease. We examined the relationship between subregional amyloid-beta and p-tau distributions in the medial…