Research
New findings, and what they do and do not show.
An MIT team knocked out every protein-coding gene, one at a time, in human oligodendrocytes carrying APOE4. The trail led to Wnt signaling, an enzyme called GSK3B, and fat droplets where insulation should be. It led there in dishes, organoids and mice, and in no patient yet.
A mathematician at Mississippi State has published a model of how metal ions drive amyloid-beta aggregation, and of what chelators and inhibitors do to that process in simulation. The paper's own claim is modest: a tool for designing experiments. The trials that tried stripping metal out of real brains are the reason that modesty matters.
A computer model of amyloid and tau biology predicts that blood tau falls by about 15 percent on one antibody and about 75 percent on another at the moment amyloid scans turn negative. Nobody was treated; every figure is a prediction.
A Mayo Clinic review of 706 memory clinic patients found that plasma p-tau217 results clashed with PET scans and spinal fluid tests in 8.5 percent of cases, and traced many of the mismatches to kidney function, body weight, sleep apnea, and how far the disease had already progressed.
A Lund University-led study of two large brain-scan collections finds that the brain reorganizes its internal traffic along two separate paths as it changes over time: one that tracks the calendar, and one that tracks Alzheimer's pathology, visible before any memory symptom appears.
In a JAMA Neurology study of more than 15,000 people screened for an Alzheimer's prevention trial, a blood marker called percent p-tau217 rose and fell with brain amyloid in roughly the same pattern for Asian, Black, White, Hispanic and non-Hispanic participants. Their baseline levels of that same marker, though, were not identical: they ran lower in Asian, Black and Hispanic participants than in White and non-Hispanic participants.
We reported the clumping enzyme. Here is the other half of the same ETH Zurich paper: small molecules that kept GRK2 from aggregating, mice that held on to more neurons and survived longer, and one result that spoils the tidy version. All of it in animals and tissue, none of it in people.
Pooling participant-level data on 38,418 people from 22 population-based cohorts in 18 countries, researchers put the number of people who developed dementia worldwide in 2023 at 11.4 million, and project 24.7 million new cases a year by 2050. Rates ran higher in women and in lower-income countries. Published in JAMA Neurology on October 12, 2026.
Trained on resting state scans from 289 older adults, a contrastive deep learning model pulled apart two tangled signals in the aging brain: one shared by everyone growing old, one specific to amyloid buildup. The amyloid signal tracked with attention and memory scores.
Researchers at the USC Schaeffer Center built a computer model of a population like the one in donanemab's pivotal trial, then asked what changes if treatment starts earlier. Their answer: $104,900 in projected lifetime value per patient, climbing 32 percent with a two year head start, and a bigger share of costs offset by lower Medicaid spending.
A national review of 1.15 million Parkinson's-related death certificates filed from 1999 through 2025 finds that the overall age-adjusted mortality rate did not change significantly over the full period, and that most of the apparent decline since 2021 traces to the fading effect of COVID-19 being listed alongside Parkinson's on death certificates, not to any improvement in underlying mortality.
A candidate-gene study of 17 experienced meditators and 17 matched controls reports blood DNA methylation differences at genes previously tied to Alzheimer's disease, including a specific APOE site, but the design cannot show whether meditation changes disease risk or course.
A multicenter study in Nature Medicine reports that serum p-tau, the protein now used in blood tests to flag Alzheimer's pathology, is also elevated in systemic amyloidosis, diseases of the heart, kidney and nerves that have nothing to do with the brain.
A study using brain tissue collected during shunt surgery has, for the first time, measured amyloid and tau turnover directly in living human brain, finding tau protein replaced within hours while amyloid plaques showed no comparable rapid turnover.
The Sydney biotech told investors its 247-person XanaMIA trial of oral Xanamem, the first long-duration test of brain cortisol control as an Alzheimer's mechanism, has finished dosing and topline results are due in November.
TrueBinding's TB006, an antibody against galectin-3, has entered a placebo-controlled Phase 2a trial in early Parkinson's disease, according to its federal registration. The approach targets neuroinflammation rather than the alpha-synuclein clumping or kinase pathways most late-stage Parkinson's drugs pursue, and no human efficacy data in Parkinson's disease exist yet.
An analysis of brain tissue from 20 people who had dementia with Lewy bodies and 20 without found no link between how many Lewy bodies had formed and how many neurons or synapses were lost, unsettling a long-held assumption about what drives decline in the disease.
A study in amyloid-positive volunteers trained machine-learning models on routine MRI and amyloid-PET measures to classify tau status, an early step toward a cheaper stand-in for tau PET imaging rather than a replacement for it.
Amyloid buildup in mice disrupted immune cell production far away in the bone marrow, blocking a supply of protective cells to the brain, according to a paper published this month in Nature Neuroscience.
Researchers profiling structural changes in plasma proteins from 520 people report a three-marker panel that distinguished healthy participants, those with mild cognitive impairment, and those with Alzheimer's disease, pointing to a biomarker approach based on protein conformation rather than concentration.
A 341-person randomized trial that told some cognitively normal older adults their amyloid, APOE and hippocampal-volume results, and delayed telling others by a year, found no difference in depression symptoms or cognitive scores between the two groups.
A multicentre trial that scanned the same 682 people with two tau PET tracers found the scans agreed on tau status most of the time but disagreed often enough to change who would count as tau-positive, a distinction that can decide trial eligibility and treatment access. Researchers say the finding does not tell them which tracer is right, since no independent test of tau in a living brain exists to check either one against.
Indiana University researchers report that deleting the Idol gene from neurons in a mouse model of Alzheimer's left far less amyloid in the cortex, and lifted a set of receptors that handle the brain's fats. It is mouse work, and it is a target, not a treatment.
A new analysis of FDA adverse-event data finds lecanemab and donanemab produce different patterns of brain-swelling and bleeding signals, the first study to compare the two anti-amyloid drugs head to head rather than separately.
A Korean team merged blood tests, MRI, amyloid scans and cognitive status into a single six-stage prognostic ladder, then tested it in an American cohort. The early rungs held. The top ones had almost nobody standing on them.
The Food and Drug Administration cleared Roche's Elecsys pTau217 blood test on Aug. 24, giving primary care and specialist clinicians a single blood draw to help assess amyloid buildup in people 55 and older with cognitive complaints. It is the fourth such blood test cleared since May 2025.
A review of 39 blood and spinal fluid studies from 2021 to 2025 found more than 180 proteins flagged as possible dementia markers. Twenty-six showed up in more than one study. One hundred and four were named by a single study and nobody else.
A Nature Nanotechnology study describes a computationally designed nanostructure, built from human ferritin cages carrying atomically precise 12-atom gold clusters, that recognizes and takes apart amyloid-beta aggregates. The work is a materials-design and mechanism paper, not a clinical trial.
A newly registered US trial will test whether an orally dosed, lab-modified relative of the spice compound curcumin can slow cognitive decline in early Alzheimer's disease, using a mechanism distinct from the antibody drugs now on the market.
A modified titration schedule for the Alzheimer's drug donanemab (Kisunla) lowered rates of a swelling side effect called ARIA-E without changing how much amyloid it cleared, according to the phase 3b TRAILBLAZER-ALZ 6 trial that the FDA cited when it updated the drug's label in July 2025.
A Nature Neuroscience study finds dopamine all but disappears from the entorhinal cortex of Alzheimer's model mice before other damage appears, and the animals' memory failures rise and fall with it. The finding comes from mice only.
A UMass Chan team applied a new segmentation tool to 654 research MRI scans and found the thalamus, nucleus accumbens and claustrum shrink in both diseases, but not to the same degree or at the same speed. It is a group-level comparison, not a test anyone can have today.
An unselected community brain collection, nearly 15,000 slides stained for tau, and seven cases. This is the counterweight to the athlete series, and it comes with hard limits of its own.
Pooling 12 population studies, researchers put crude incidence of dementia with Lewy bodies at 4.79 per 100,000 person-years, climbing steeply after 65. The wide confidence bands are part of the finding.
A Neuron paper from Stanford and UCSF describes a route from tau to neuron death that never passes through a tangle. It is fly, mouse, stem cell and post-mortem work, and the compound that blocks it has not been given to a single person.
A new Frontiers in Neuroscience analysis of 17 years of Brazilian Ministry of Health records finds 1,576 suspected Creutzfeldt-Jakob notifications where the authors calculate about 3,200 cases should have occurred. A Mexico City emergency department series points the same way from the opposite direction.
In a mouse study published on 5 February 2026 in the Journal of General Virology, cofactor-free prion protein fibrils induced misfolded protein and spongiform brain damage in knock-in animals carrying the E200K mutation, but not in animals with wild-type prion protein.
A large unrefereed preprint asked whether clinical notes, MRI and spinal fluid can predict what a pathologist will find. Alzheimer's staging came through. Lewy bodies, TDP-43 and vascular injury, the things that make a brain mixed, did not.
A Swiss study of 363 community-dwelling adults reports myelin loss and fluid at the core of white matter hyperintensities, with axonal change in the tissue around them. It is imaging only, at a single time point, with no tissue examined.
An Oxford calculator estimates multidomain cognitive impairment six months after stroke; an international consortium model estimates five-year dementia risk. Neither has been tested as a decision aid in clinical practice.