AI Longevity Daily

Today’s Summary

Sickle cell gene therapy directly corrects the disease-causing mutation in early human trials, achieving transfusion independence.
Blood sample research will track early protein signals before dementia symptoms appear — still exploratory analysis.
The gene therapy has early data from only three patients; no efficacy conclusions can be drawn yet.

🔥 Top 10 Today

1. Sickle Cell Gene Therapy Directly Repairs the Mutant Gene in 3 Patients

Kamau Therapeutics has dropped early data from its RESTORE Phase 1/2 trial: all 3 patients treated with nula-cel achieved transfusion independence, with zero pain crises post-treatment. The approach uses homology-directed repair (HDR) — think of it as swapping out the “typo” in a gene for the correct version, rather than working around the problem by activating fetal hemoglobin the way the already-approved Casgevy does. In the two patients who completed the 6-month follow-up, disease-causing sickle hemoglobin (HbS) dropped below 20% of total hemoglobin, while normal adult hemoglobin (HbA) hit 77% and 67% respectively. As GEN reports , this is early clinical proof that HDR can directly correct the native β-globin gene in real patients. Full data drops December 14 at the ASH annual meeting — but with only 3 patients, no efficacy conclusions yet. 🧬

2. UK Biobank’s UKB-PRISM Study Tracks Early Protein Signals of Dementia in 80,000 Blood Samples

UK Biobank’s UKB-PRISM study plans to measure hundreds of circulating proteins across 80,000 stored blood samples, hunting for biomarkers of Alzheimer’s and Parkinson’s that appear years — even decades — before symptoms show up. The testing will run on Alamar Biosciences’ NULISA ultra-sensitive protein detection platform at Nightingale Health labs, with completion expected by end of 2027. As Longevity.Technology reports , these samples were collected before participants got sick and are linked long-term to genetic, imaging, and medical records — meaning protein findings can be matched against real subsequent diagnoses. Caveats: protein levels are correlated with disease but can’t yet be used for individual diagnosis, and the UK Biobank population skews healthier and wealthier than the general public. 🧠

3. NOVOS Core Supplement RCT: Vascular Markers Improve, Lipids Don’t Budge

A University of Surrey team ran a double-blind RCT on 61 healthy adults over 40, taking either a 12-ingredient NOVOS Core supplement or a placebo daily for 6 months. NOVOS Core delivered some solid vascular wins: flow-mediated dilation (FMD — a measure of how well blood vessel walls expand) improved roughly 47% relative to baseline in the supplement group, arterial stiffness dropped, and systolic blood pressure came in 6.1 mmHg lower than placebo. Lipid markers? No change. As Longevity.Technology reports , the study is published in the European Heart Journal Open, but only 43 participants finished, there’s no way to tell which of the 12 ingredients did the work, and the final measurements were taken about 24 hours after the last dose — so whether the gains stick around is still an open question. 💊

4. Biogen’s Tau-Targeting Drug Diranersen Will Unveil New Data at Alzheimer’s Conference

Biogen is set to release interim results for diranersen at the CTAD conference in Boston on November 17. Diranersen is an antisense oligonucleotide (ASO) — basically a “shut-down instruction” that tells cells to produce less tau protein. Tau buildup inside neurons is a core pathology of Alzheimer’s and is directly tied to nerve cell damage. The Phase 2 CELIA trial enrolled 416 patients with mild cognitive impairment or mild dementia over 18 months, and will present 96-week long-term extension data alongside tau biomarker results. As Longevity.Technology reports , the full early efficacy numbers haven’t been made public yet, and whether the results hold up over a longer time window remains to be seen. 👀

5. Multi-Objective Molecule Generation Fine-Tuning: Balancing Binding Activity and Synthesizability Simultaneously

HVFT (Hypervolume Fine-Tuning), proposed in a preprint (not yet peer-reviewed), steers a single pretrained generative model to optimize multiple conflicting molecular properties at once — like designing a molecule that both binds its target and is easy to synthesize in the lab. The key move is converting “expected hypervolume” — a metric capturing overall coverage across multiple properties — into a per-candidate adaptive reward signal that works with standard training pipelines. The paper index/abstract shows it outperforms existing baselines on therapeutic peptide and small molecule design tasks. Still a preprint, validated only on benchmark tasks, with no wet lab results yet. ⚗️

6. FoldScape Predicts Protein Folding Intermediates Across the Full Journey from Unfolded to Near-Native

Understanding how a protein folds from a floppy chain into a functional 3D structure matters enormously for disease biology and drug design. FoldScape, proposed in a preprint, is a deep learning framework that generates conformational ensembles at user-specified folding progress levels given a sequence — basically taking “snapshot segments” of the folding process rather than just handing you the final structure. Tested on 20 human proteins, FoldScape covers roughly 10× the folding coordinate range of AlphaFold 3 multi-seed sampling, though it’s less accurate near the native state — making the two complementary. The paper index/abstract notes this is a preprint, not yet peer-reviewed, and not experimentally validated. 🔬

7. E2I-Flow Reconstructs Visual Images from EEG Brain Signals

E2I-Flow proposes a framework for decoding electroencephalography (EEG — the brain’s electrical activity recorded via scalp electrodes) into visual images. The method uses dual-path EEG feature extraction with wavelet frequency and temporal mixing branches, then generates corresponding images via a conditional flow matching model. The paper index/abstract shows it beats multiple baselines on zero-shot image retrieval and reconstruction tasks. Still a preprint, performance depends on public benchmark datasets, and no clinical application validation exists yet. 🧩

8. Sleep Quality Tied to Brain Health in People with Cardiometabolic Disease

Research from the Karolinska Institute finds that among people with type 2 diabetes, heart disease, or a history of stroke, those with better sleep quality show sharper cognitive function and larger brain volume. As News-Medical reports , specific sample size, follow-up duration, and study design details are pending verification. Based on available abstract info, this is a correlational study — it can’t conclude that improving sleep causes brain volume to increase. 😴

9. Eating Peanuts Daily May Alter the Brain’s Insulin Response

A study finds that daily consumption of unsalted peanuts may enhance the brain’s insulin sensitivity, offering a new clue about how diet affects cognitive health in aging. The brain’s insulin response is linked to memory and cognitive function, and declining sensitivity has been associated with cognitive deterioration. As News-Medical reports , specific sample size, intervention duration, and measurement methods are pending verification — this can’t be used to conclude that eating peanuts prevents cognitive decline. 🥜

10. Molecular Prior Networks: One Forward Pass for Drug-Target Affinity Prediction, No Task-Specific Retraining

In drug discovery, predicting how strongly a small molecule binds its target protein usually means retraining a model from scratch for each new target. Molecular Prior Networks sidestep this by pretraining a Transformer across thousands of existing affinity datasets, so at prediction time you only need a single forward pass — no retraining required. The paper index/abstract shows it hits state-of-the-art on FS-Mol benchmark low-support scenarios, making it a natural fit for data-scarce early-stage drug discovery. Still a preprint, no wet lab validation yet. 💡

⚡ Product & Feature Updates

Full-Season Wearable Tracking Program Launches for Adolescent Female Athletes

As Longevity.Technology reports , all drivers and coaches in the More than Equal racing driver development program are now wearing the Ultrahuman Ring for full-season tracking of sleep, heart rate variability (HRV — a reflection of how well the autonomic nervous system regulates recovery, which trends downward with age), and daily recovery scores. Most of these athletes are still teenagers — and peak bone mass, cardiorespiratory fitness, and endocrine rhythms are primarily established during adolescence, making this a critical window for long-term health. Each driver’s readings are compared against their own baseline rather than population averages, and menstrual cycle tracking is managed separately by the team. This partnership doesn’t publish data externally or feed into formal research — and Ultrahuman Ring’s daily recovery scores are not equivalent to biological age measurements.

🧪 Frontier Research

SphereFlow Closes 80% of the Gap to Full Multimodal MRI for Missing Sequence Completion

Brain tumor MRI typically requires multiple sequences (T1, T2, FLAIR, etc.), and clinical conditions often mean one sequence goes missing. SphereFlow projects all modalities into a shared spherical latent space and uses flow matching to complete cross-modal gaps — rather than training a model to make a “long leap” between two modalities. The paper index/abstract shows that on the BraTS brain tumor multimodal MRI dataset, using SphereFlow-completed images for downstream segmentation closes about 80% of the performance gap versus having all four complete modalities as input. Code is open-sourced at github.com/WTHH031230/SphereFlow . Still a preprint, not yet validated in clinical settings. 🏥

⌘ Top Open-Source Projects

qlu-lab/SARDINE: Estimating Epigenetic Clock–Phenotype Associations from GWAS Summary Statistics

SARDINE is a statistical framework for analyzing associations between epigenetic clocks (tools that estimate biological age via DNA methylation patterns) and health phenotypes. It leverages epigenome-wide association study (EWAS) summary statistics to estimate effect sizes without needing access to raw individual-level data. Ideal for researchers who want to draw inferences about biological age and disease relationships even when raw methylation data isn’t available. Repository documentation and license details can be confirmed on the GitHub page. Stars: 1, last active: 2026-11-03.

NoWon1/AI-challenge-for-HEALTHY-Brain-Aging: Multimodal Prediction Models for Dementia Risk and Progression

This repository is competition code for the CBR Healthy Brain Aging AI Challenge, integrating multimodal data to build dementia risk and progression prediction models, including a “lightweight digital twin” component that attempts to model individual brain health states. Good reference for researchers interested in brain aging prediction model design and challenge solution architecture. The repo description doesn’t provide specifics on input formats or dependency environments. Stars: 1, last active: 2026-11-04, runtime status unverified.

❓ Related Questions

Do the NOVOS Core Trial Results Mean the Supplement Actually Extends Lifespan?

Not yet. The trial measured proxy indicators of vascular function (FMD, arterial stiffness, blood pressure) — these metrics have statistical associations with cardiovascular risk in large population studies, but a 6-month trial with 43 completers can’t support a “extends lifespan” conclusion. The study itself acknowledges it can’t distinguish which of the 12 ingredients drove the effect, and can’t rule out confounding from the placebo group having a higher baseline age. Longevity.Technology’s coverage and the original paper (European Heart Journal Open, doi:10.1093/ehjopen/oeag157) have the full methodology and limitations. 🔍