The limiting factor in this disease is not model capability. It is that the datasets a method needs are small, held by few centres, and in most cases not shared at individual level. This page states what can be obtained and on what terms, so that a proposal is scoped against reality. Problems worth solving covers which questions are worth the effort.
What is openly available?
Three things, and they are enough for a literature or registry project but not for model training on patient-level data.
Trial registry records. Design, arms, endpoints, sites, status and dates are public and machine readable, and changes over time are visible in record histories. Those records are the basis of most of what this site states about trials.
The published literature. Abstracts, metadata and open-access full text for prion disease are retrievable through the main bibliographic services, and much of the clinical description for rare variants exists only as case reports there.
Aggregate surveillance figures. Published national summaries give counts and proportions. The Canadian surveillance factsheet, for example, reports 90 people diagnosed in 2022, about 90 percent of diagnosed people aged 45 to 85, incidence of about 1 to 2 cases per million people per year, and about 6 percent of cases inherited, about 93 percent sporadic and under 1 percent acquired. The page does not describe how researchers can access underlying data, and gives notification contacts only.
What about individual-level data?
Mostly closed, with one documented route.
The PrProfile trial of ION717, a phase 1 and 2 study with 85 participants enrolled, states that the sponsor may share anonymised individual participant data, and that requests from qualified researchers will be considered once stated criteria are met. That is a request process with conditions, not a download.
The PROSPECT registry, a prospective observational study of adults at risk for genetic prion disease run from University Hospitals Cleveland Medical Center with Case Western Reserve University, the CJD Foundation and Prion Alliance as collaborators, is recruiting United States residents with annual assessments for up to 10 years and a target of about 2,000 participants. Its record states that individual participant data sharing is not planned.
Those two records bracket the field: industry data behind a request process, academic cohort data not offered.
What will a developer not find?
No public patient-level imaging, cerebrospinal fluid biomarker or assay dataset for human prion disease appears in the sources read here. Where published studies report assay sensitivities, the underlying sample-level values generally sit with the originating centre.
Brain tissue is held in biobanks rather than released as data, and access is through application to the holding institution.
Population allele frequency resources are an exception worth naming, because the penetrance estimates this field relies on were produced by comparing patient variants against large population control datasets. That comparison was possible precisely because the control data was open.
What this means for a proposal
Scope to the registry and literature layer first, and treat anything needing individual records as requiring a named collaborating centre and an application, with a timeline measured in months rather than weeks.
Expect small numbers. A variant group in the largest multicentre biomarker cohorts can be a handful of cases, and a method that needs thousands of labelled examples has no path here.
Expect privacy constraints that are stricter than the headline rules, because a cohort of carriers of a named variant in one country can be identifiable from few fields. Genetic data and privacy rules sets out the legal frame.
We hold none of this data ourselves and do not broker access. Responsible use states what we ask of tools that use our pages, datasets lists what we have catalogued, and corrections with a public source are welcome at corrections@thegenerativebrain.org.