The core that decides who gets modeled

What the work claims

This is an active core-infrastructure grant, not a results paper; the record describes a designed service, so every scientific claim in it should be read as intent rather than finding. The core, led by Jerome S. Mertens at UC San Diego and funded at $462,945 in FY2026 within a center award of $4,584,783 that year, exists because the record's premise is that human cell models, including induced pluripotent stem cell (iPSC) models, are essential to understanding Alzheimer's disease and related dementias, and that after more than a century of study no treatment can prevent, stop, or reverse progression. The core's five aims run from banking to distribution: keep deriving iPSC lines from participant skin fibroblasts with rigorous quality control, cataloging, and banking for prolonged sharing; build and longitudinally follow a Latino participant cohort to diversify the biorepository, on the stated rationale that this population is more likely to develop Alzheimer's yet remains understudied; offer isogenic, meaning genetically matched except for one engineered edit, iPSC lines generated on demand, plus organoid differentiation services and training; stand up an observatory of induced neurons, iNs, converted directly from participant fibroblasts, described as the aging-relevant model; and run consulting and hands-on training across the portfolio.¹

How the pipeline works

The core is a cohort-to-platform converter. The center's Clinical Core recruits participants, including the longitudinal Latino cohort. Skin fibroblasts are isolated, reprogrammed into iPSCs, which are adult cells reset to a stem-like state capable of differentiating into many cell types, then quality-controlled, cataloged, and banked so they can be shared for prolonged periods. Downstream, the core operates three service tiers. Isogenic editing on demand: a researcher names a variant of interest and receives an engineered line paired with its unedited control from the same donor, which isolates the variant's effect from the donor's genetic background. Organoid differentiation services with training: labs that lack the craft get finished neural organoids plus the skill transfer to make their own. And the iN observatory: neurons made by direct conversion, skipping the reprogrammed stem-cell state, which the record presents as the aging-relevant option, since the route is chosen specifically to study neurodegeneration as it presents in aged donors.¹

Each tier answers a different access failure. Banking fixes availability, isogenic pairs fix genetic confounding, differentiation services fix technique, and the observatory fixes a model-aging mismatch. The design is coherent; what it is not is neutral, because every tier also encodes choices about whose biology becomes the reference.

Where a skeptic should push

The load-bearing assumption is that a banked, shared iPSC collection from this cohort will generalize to Alzheimer's biology in the broader population. The record asserts the disease-modeling rationale but contains no results, no sample sizes, no power considerations, and no stated validation of the Latino cohort's downstream model fidelity. A representativeness fix at recruitment only works if enrollment succeeds and attrition stays low across a longitudinal design, and longitudinal cohorts skew toward participants with access, stability, and trust. The cohort can be more representative than the status quo and still not representative.

Second, the aging claim deserves scrutiny. Direct-conversion iNs are presented as aging-relevant without any supporting data in the record; the reasoning that reprogramming can erase age-associated cellular signatures while direct conversion preserves them is plausible and grounded in the field's published literature, but this record neither demonstrates nor cites it, so it stands as motivation, not evidence. Third, the consent horizon: participants consent to fibroblast donation in a research context framed around an Alzheimer's center, while the core's own language promises prolonged sharing, on-demand genome editing, organoid differentiation, and distribution of materials and tools to the research community. Whether the consent instruments cover derivative lines, edited lines, and distribution to unspecified future investigators is exactly the kind of question a core of this scope should answer in public, and the record is silent on it. Finally, on-demand editing raises a quieter question: the core decides which edits are feasible and worth making, which means a service organization is participating in setting the field's research agenda.

Who gets built into the platform

The access lesson cuts in both directions, and this core illustrates both. On the opportunity side, the service tiers are a genuine democratization mechanism. Isogenic pairs, finished organoids, training, and a shared observatory lower the skill and capital floor for labs that are not stem-cell boutiques, and an explicit, funded commitment to an understudied population pushes back on the convenience-sampling drift that makes most cell-model libraries demographically narrow. If a disease-modeling platform is infrastructure, then representativeness is an engineering requirement, and writing it into an aim statement is the right move.

On the threat side, the same infrastructure concentrates quiet authority. Whichever donors enter the bank define the reference genetic backgrounds against which variants will be interpreted for years; the cohort chosen in 2019 becomes the implicit standard for what Alzheimer's looks like in a dish through 2029 and beyond. Consent architecture built for a local research biobank may not stretch to cover prolonged, editable, distributable derivatives a decade later, and the record does not say how, or whether, participants are re-engaged as the service menu widens. There is also a dependency hazard: a field that routes its baseline models through one core inherits that core's funding cycles, its banking decisions, and its outages. Access widened through a single spigot is access on the spigot owner's terms. The governance question this core poses is not whether it does good work; it is whether the decisions that make the platform representative, and the decisions that make it distributable, ever get reviewed as the governance decisions they are.

The bottom line

Designed, not demonstrated. As infrastructure, the core is well-conceived: it attacks real access failures, and its representativeness aim is more explicit than most. As science, nothing here is yet a finding, and the two claims that matter most, that the cohort will yield generalizable models and that the iN observatory captures aging biology, are motivations awaiting data. What would confirm the design: publications from distributed users showing the banked and edited lines reproduce known Alzheimer's biology and that direct-conversion iNs from aged donors carry age-associated signatures that iPSC-derived neurons lose. What would break it: high longitudinal attrition in the cohort, or downstream models that fail to recapitulate disease phenotypes despite the infrastructure investment. Either outcome would tell the field something valuable about whether representativeness can be engineered at the core layer.