Alzheimer cfDNA application pack (example)¶
Instance of the methylation application-pack pattern
(disease application). A staged (Control -> MCI -> AD) cfDNA DNA-methylation study
on the existing methylation SaMD control plane. It selects assay procedure
cfdna_wgbs_plasma (linear WGBS, fragmentomics, gene FeatureCuts, no cell
deconvolution), the samd_* profile ladder, and the cfDNA analyte profile. The pack
is config + cohorts + partitions + a disease overlay — no new actions or aligners.
See the operator guide: Usage ch.21 Alzheimer cfDNA pack and the shared methylation workflow: end-to-end workflow.
Files¶
| File | Role |
|---|---|
project_Healthy_vs_AD_Stages.json |
Study manifest: Control vs MCI vs AD stages, cfDNA analyte, methylation modality, progression labels, partitions (placeholder IDs) |
context_alzheimer_cfdna.json |
Disease overlay: pipelineProcedure=cfdna_wgbs_plasma, mapper.disease_term=Alzheimer's disease, enricher.library_preset=neuro-core, progression |
data/*.csv |
Cohort CSV stubs (replace placeholder IDs with real, patient-disjoint sample IDs) |
Instantiate on /work¶
Studies live under /work/projects/<study>/. Scaffold with the CLI, then copy this
manifest/overlay in:
methyl-study-init \
--study-id alzheimer-cfdna \
--name Healthy_vs_AD_Stages \
--analyte cfdna \
--modality methylation \
--stages 2 \
--intended-use "Research-use cfDNA methylation classifier for Alzheimer detection." \
--output-root /work/projects
# then rename the two disease stages to MCI / AD (see this manifest), fill data/*.csv,
# and assign patient-disjoint validation_partitions.
Run the SaMD ladder¶
# Research (procedure points at no-deconv lifecycle)
methyl-workflow-run \
--program workflow_engine/domain/fixtures/study_validation_lifecycle_no_deconv.program.json \
--context-file docs/examples/samd/alzheimer-cfdna/context_alzheimer_cfdna.json
# Enrichment (requires non-empty locked_test) -> switch pipelineProfile to samd_holdout_enrichment
# Pivotal (requires non-empty pivotal_validation) -> switch pipelineProfile to samd_pivotal
Enrichment uses the neuro-core library preset from the enrichment preset registry
(methyl-cfg sync-library-presets), not the oncology cancer-core default.