The Biggest Alzheimer’s Study You Haven’t Heard About — And the Question It Forgot to Ask

They found 468 proteins linked to Alzheimer’s disease that nobody had seen before.

And they may have missed half the picture.

A study published this month in Advanced Science did something deceptively simple that changed what we can see about Alzheimer’s disease. Instead of lumping all AD patients together and adjusting for APOE genotype — the way virtually every prior proteomic study has done — Li et al. separated their 3,060 participants by APOE genotype and analyzed each group independently.

The result was staggering: 64–78% of the proteins they identified had never been found in non-APOE-stratified studies. Proteins that had been invisible for years — hidden by the statistical noise of treating APOE as a covariate rather than a biological variable — suddenly appeared.

This is not a minor technical point. It is a methodological lesson with implications far beyond this single study. When you adjust for a variable, you assume it shifts the baseline. When you stratify by it, you allow it to change the biology. Li et al. proved that APOE doesn’t just shift the baseline — it changes which proteins are associated with Alzheimer’s disease in the first place.

What they found when they looked is worth paying attention to.

What the APOE-Stratified Lens Revealed

The proteins and metabolites associated with AD in APOE ε3/ε3 carriers (the “average” genotype) and APOE ε3/ε4 carriers (the single-copy risk genotype) told a story of converging damage:

Mitochondrial dysfunction was the dominant signal — fatty acid β-oxidation failure, TCA cycle disruption, branched-chain amino acid degradation collapse. The mitochondrial matrix was the most significantly enriched pathway among proteins that were downregulated in AD across both genotype groups.

Lipid accumulation was pervasive — and the pattern revealed something unexpected. When Li et al. compared AD patients to controls within the same APOE genotype group — ε3/ε4 cases against ε3/ε4 controls, ε3/ε3 cases against ε3/ε3 controls — the AD group still showed greater lipid dysregulation. Both sides of each comparison carry the same APOE alleles. The genotype is held constant; only disease status differs. So whatever is driving the additional lipid accumulation in the people who actually developed Alzheimer’s isn’t APOE itself — it’s something else. Something that APOE4 may set in motion but that then takes on a life of its own.

The nature of the lipids that accumulated offers a clue. Lysophospholipids — the products of phospholipase A₂ activity — were the dominant upregulated lipid class. Phospholipase A₂ is the enzyme that cleaves fatty acids from the sn-2 position of membrane phospholipids. When you see lysophospholipids accumulating, you are seeing the molecular debris of membrane remodeling — evidence that something is actively stripping fatty acids from neuronal membranes. In APOE4 carriers, cPLA₂ (the calcium-dependent form) is chronically overactivated, preferentially liberating arachidonic acid from sn-2 positions. The lysophospholipid accumulation Li et al. found is the downstream signature of this process — the empty phospholipid scaffolds left behind after cPLA₂ has done its work.

But here’s the question the study doesn’t ask: if fatty acids are being stripped from membrane sn-2 positions, what is replacing them?

The answer matters more than the loss itself. When DHA is removed from the sn-2 position of phosphatidylethanolamine — the dominant phospholipid in neuronal membranes — the Lands cycle doesn’t leave the position empty. It refills it. And in the context of APOE4-driven cPLA₂ overactivation and omega-6 excess, the replacement species is increasingly adrenic acid — a 22-carbon omega-6 fatty acid that happens to be the preferred substrate for ferroptotic cell death.

That replacement — and why it may matter independently of DHA loss — is the guaranteed subject of my forthcoming framework and clinical offering.

The APOE4-Specific Proteins Nobody Had Seen Before

Li et al. found that 86% of AD-associated proteins showed similar effect sizes in ε3/ε3 and ε3/ε4 carriers — meaning most of the proteomic damage in AD is shared across genotypes. But 17% of proteins were uniquely dysregulated in one genotype, and these told a distinct story.

CPLX1 — complexin 1, essential for synaptic vesicle exocytosis and neurotransmitter release — showed the largest effect size difference between genotypes. CASQ1 and CRACR2A, both involved in calcium regulation, were elevated specifically in ε3/ε4 carriers. GSTP1, a glutathione S-transferase that protects cells from oxidative stress by conjugating glutathione to toxic substrates, was also APOE4-specific. None of these had been reported in any previous AD proteomic study.

Together, they paint a picture of a cell under a very specific kind of stress: membrane dysfunction driving calcium dysregulation, synaptic vesicle trafficking failure, and a desperate upregulation of detoxification machinery. The cell is not just inflamed — it is trying to survive an oxidative assault that originates at the membrane.

The study also performed drug repositioning analysis, identifying FDA-approved compounds that might target the uncovered signaling networks. This is where the story gets both exciting and dangerous — because of the question they forgot to ask.

The Question They Forgot to Ask

Li et al. acknowledged one major limitation in their Discussion: “We did not analyze if any proteins or metabolites were sex or age specific. Recent studies have identified proteomic signatures specific to males or females. Therefore, further analyses stratifying by sex and APOE will be needed to fully disentangle the different pathways implicated in AD.”

This is not a minor caveat. It is the same methodological gap they just proved matters for APOE — applied to a variable with equally strong biological justification.

Consider the logic: Li et al. demonstrated that adjusting for APOE as a covariate misses 64–78% of AD-associated proteins. They argued — correctly — that stratification reveals biology that covariate adjustment conceals. But they adjusted for sex as a covariate in every analysis. If sex modifies the AD proteomic signature the way APOE does — and the evidence says it does — then a comparable percentage of sex-specific proteins may be hiding in their data right now.

The evidence that sex stratification would reveal different biology is not speculative. It is published, replicated, and increasingly urgent.

Arnold et al. (2020) conducted exactly the kind of sex × APOE stratified metabolomic analysis that Li et al. did not — in 1,517 ADNI participants. They found metabolic effects limited to APOE ε4-positive females that were invisible in every other analytical framework. Their conclusion: females experience greater impairment of mitochondrial energy production than males in AD.

Maffioli et al. (2022) used integrated proteomics and metabolomics in human hippocampus and found that AD “almost reverses” sex-specific molecular profiles — insulin and serine metabolism showed fundamentally different patterns in males versus females.

Do et al. (2026) found that male-specific AD proteins were enriched in microglia and involved in activating innate immune response, while female-specific proteins were enriched in endothelial cells and involved in protein metabolic regulation — fundamentally different cellular origins.

Comas-Albertí et al. (2026) used NULISAseq across AD, Lewy body dementia, and FTD and found that sex-stratified plasma analyses revealed substantially more differentially expressed proteins than unstratified analyses, with females showing a higher number of upregulated inflammation-related proteins.

And then there is the finding that should stop every AD drug developer in their tracks. It compels me to look at every study I read with a very specific, sex-based lens…

The Delivanoglou Discovery: When the Same Drug Helps Women and Hurts Men

Delivanoglou et al. (2026), published in Neuron, demonstrated that APOE4 has sexually dimorphic effects on brain lipid composition and neuroinflammation. That alone would be important. But the critical finding was this: suppressing innate immunity via CSF1R inhibition restored cognition in E4/E4 females while exacerbating cognitive decline in E4/E4 males (mice model).

The same intervention. Opposite outcomes. Determined entirely by sex.

Li et al. identified complement and coagulation cascades (C3, C4A, C5, CLU) as upregulated in AD across APOE genotypes. They performed drug repositioning to find compounds targeting these pathways. But that drug repositioning was done on sex-pooled data. If the Delivanoglou finding generalizes — and Shi et al. (2025) showed that drug repositioning on sex-stratified transcriptomic data reveals drugs with potential opposite effects in different sexes — then some of the therapeutic candidates Li et al. identified may be beneficial for women and harmful for men, or vice versa.

This is not a theoretical concern. It is a concrete risk embedded in the study’s drug repurposing pipeline.

The Fatty Acid Story Gets Stranger — and More Sex-Specific — Than Anyone Expected…