Molecular Pathology Variant Curation Workflow Chain Specification

Role: Lead Bioinformatics Workflow Engineer specialising in clinical genomics automation pipelines.

Context

• Sequencing Assay: {{sequencing_panel_type}}
• Knowledge Base Endpoints: {{evidence_database_endpoints}}
• Classification Framework: {{acmg_classification_criteria}}
• Hallucination Guardrail Protocol: {{hallucination_guardrail_method}}
• Molecular Pathologist Handoff Rules: {{pathologist_handoff_rules}}

Task

Design a multi-stage automated workflow chain that ingests annotated variant call files from {{sequencing_panel_type}}, gathers peer-reviewed evidence via {{evidence_database_endpoints}}, computes candidate pathogenicity under {{acmg_classification_criteria}}, and outputs structured curation dossiers following {{pathologist_handoff_rules}}.

Method

1. Specify the variant normalization node that converts incoming variant call formats from {{sequencing_panel_type}} into standard HGVS nomenclature.
2. Design the query dispatch agent to parallelize evidence fetching across {{evidence_database_endpoints}} with rate-limiting and cache controls.
3. Formulate the literature extraction agent to isolate functional assay data, population allele frequencies, and clinical case reports.
4. Integrate {{hallucination_guardrail_method}} to verify every cited PMID, functional assay claim, and cohort statistic against primary source text.
5. Construct the rule-based aggregation step that evaluates evidence strength against the specific rules of {{acmg_classification_criteria}}.
6. Detail the draft report synthesis agent that compiles tiered molecular summaries and therapeutic matching options.
7. Define the clinical handoff state machine governing review priority based on {{pathologist_handoff_rules}}.

Constraints

• MUST ground every classification criterion in a validated primary literature identifier or database accession number.
• MUST NOT assign final pathogenic or benign status without applying {{hallucination_guardrail_method}} verification.
• All intermediate JSON outputs must conform to strictly defined biological ontology schemas.
• Missing or conflicting literature evidence must be explicitly surfaced as high-priority flags in the summary dossier.

Output format

Provide the technical specification organized into four sections:

1. Pipeline DAG & Agent Architecture: Node diagram and execution sequencing rules.
2. Agent Step Specifications: Detailed inputs, knowledge base query templates, and output validation criteria.
3. Verification & Guardrail Engine: Implementation mechanics of {{hallucination_guardrail_method}}.
4. Clinical Report Specification: Structured output schema delivered to the reviewer per {{pathologist_handoff_rules}}.

Self-review

• Confirm all 5 variables ({{sequencing_panel_type}}, {{evidence_database_endpoints}}, {{acmg_classification_criteria}}, {{hallucination_guardrail_method}}, {{pathologist_handoff_rules}}) are thoroughly integrated.
• Ensure no agent step allows hallucinated PubMed IDs or synthetic variant descriptions to propagate downstream.
• Check that the evidence grading logic strictly conforms to the specified ACMG/AMP ruleset.