Genomic pipelines a clinician can sign, and an auditor can replay.

Metagenomics, isolate genomics and RNA-Seq run their own rails with the same discipline: QC before compute, pinned references, a review gate, and a provenance record that lets any run be replayed. That is what an accreditation assessor asks for, and what the validation report is written against.

1. Reads (Intake). Raw reads from Illumina, MGI or Oxford Nanopore, checksummed on arrival and registered against the sample. Tools: FASTQ, Checksums, Sample manifest.
2. QC (Quality control). Read quality, contamination and coverage checked before any compute is spent on the sample. Tools: FastQC or equivalent, Coverage thresholds per assay. Gate: A sample below the coverage threshold stops here.
3. Align and call (Secondary analysis). Alignment and variant calling under GATK Best Practices, GPU-accelerated where the science permits. Tools: GATK, NVIDIA Parabricks, AWS Batch, Azure Batch or Cloud Batch.
4. Annotate (Annotation). Variants annotated against pinned reference and database versions, recorded per run so the annotation can be reproduced. Tools: ClinVar, Pinned reference genome, Version-pinned databases.
5. Classify (Interpretation). Evidence assembled per variant into a draft ACMG/AMP classification for a clinician to review. Tools: ACMG/AMP criteria, Evidence per variant. Gate: A clinician signs before a report exists.
6. Report (Clinical output). A tiered clinical report aligned to NABL workflows, exportable as FHIR for ABDM interoperability. Tools: Tiered report, FHIR resources.
7. Provenance (Audit). Every tool version, reference, parameter and decision recorded, so the run can be replayed and the report defended. Tools: Nextflow, Docker, Tamper-evident audit trail.