Skip to content

SaMD submission scaffold (510(k) / De Novo map)

Disclaimer: This document is an internal documentation scaffold. It maps FDA-style submission topics to existing MethylPipeline controls and evidence. It is not a filed 510(k) or De Novo, and does not constitute FDA clearance or approval. Architecture, profiles, and filled feasibility evidence packages ≠ regulatory authorization.

Purpose

Help quality and clinical teams assemble a submission narrative by pointing each content area at canonical repo artifacts and on-disk study evidence.

Operator study path: ../usage/18-samd-study-lifecycle.md.
Evidence registry: validation-evidence-index.md.

Claim gate (code-enforced)

Requirement Where enforced
regulatory.stage must be pivotal_validation (or later) before clinical performance claims RegulatoryLifecycleConfig in packages/methylvalidation/methyl_validation/config.py
allow_clinical_performance_claims: true forbidden on earlier stages same
Real holdouts for enrichment/pivotal profiles methyl-study-validate-manifest + study validation_partitions

Current prostate packages in the evidence index are feasibility with no partitions — they must not be cited as pivotal clinical performance.

Content map (scaffold)

Submission topic (illustrative) MethylPipeline control / artifact
Device description / intended use Study regulatory.intended_use_summary, claim_boundary, analyte; SaMD SOP ch.18
Software architecture / SoTA ../architecture/index.md, methylpipeline-product-and-operational-controls.md
Design controls / change control change-management-plan.md, traceability-matrix.md
Configuration management Four-layer config (site / profile / study / program); config-not-code; CAAS / hyperparamSetId
Verification (unit/integration) continuous-integration-and-regression-testing.md, CI JUnit
Deployment / cybersecurity deployment-and-supervision.md, worker/gateway env templates
Analytical validation (stability, freeze) monte_carlo_runs/stability/, freeze readiness AR, production locked_model_spec.json
Clinical performance (when allowed) clinical_performance_report.json (Wilson/LCB); WF3 holdout eval; requires pivotal stage + partitions
PCCP / post-market pccp_draft.*, post_market_monitoring_scaffold.* from regulatory_artifacts.py (draft scaffolds)
Traceability of a specific result Evidence package in validation-evidence-index.md

Prostate feasibility packages (starting points)

ID Study Use in a draft submission narrative
EV-PCA-PLASMA-2026-06 Plasma_healthy_vs_PCa Richest chain (stability → production → PMV clinical reports); still feasibility
EV-PCA-HGOOD-2026-06 H_PCa_good Freeze/model present; gene panel empty at threshold — document limitation
EV-PCA-BUFFY-2026-07 Buffy_healthy_vs_PCa Stability + readiness only; incomplete model chain

SaMD profile ladder

Operator profiles on the claim path (see also ch.18):

Profile Role
samd_research Discovery / feasibility; research-mode overlays when folded
samd_holdout_enrichment Locked holdout enrichment; WF3 on locked_test
samd_pivotal Pivotal validation; clinical performance claims only after stage + review
  1. Scaffold study with methyl-study-init; assign patient-disjoint locked_test early.
  2. Research with samd_research (+ research mode overlay when folded); lock HPs.
  3. Enrichment with samd_holdout_enrichment; WF3 on locked_test.
  4. Open pivotal_validation cohort; run samd_pivotal; set stage + claims only after review.
  5. Fill a new evidence package (not feasibility drafts) with release SHA, CI, and partition IDs.