Design Genomics Data Pipeline - System Design | WinJob

Genomics pipeline framing: from sequencer to clinical report

Design a genomics data pipeline (Illumina / 23andMe / DNAnexus class): ingest terabase FASTQ from sequencers, run reproducible alignment and variant calling on HPC, and release DUO-gated CRAM/VCF to analysts. This section focuses on genomics pipeline framing: from sequencer to clinical report in the understanding phase—not a generic batch ETL lecture.

Numbers to state early

Metric A: 120 runs/day
Metric B: 30x WGS depth
Metric C: GxP audit trail

Mechanism

Instrument uploads demultiplexed FASTQ to a landing zone; LIMS binds sample_id, study cohort, and consent flags before any compute job is admitted.

Failure and edge cases

Treating genomics like a nightly ETL job ignores reproducible workflow versions, reference genome builds, and partial reruns from arbitrary pipeline stages.

When discussing Genomics pipeline framing: from sequencer to clinical report, trace Sequencer → LIMS → Orchestrator with 120 runs/day and explain how workflow pinning, reference builds, and consent/DUO gates differ from ad-hoc Spark jobs.

Java

javaOne Dark Pro

1 public record SampleManifest(String sampleId, String studyId, String referenceBuild) {}

Python

pythonOne Dark Pro

1 @dataclass(frozen=True)
2 class SampleManifest:
3     sample_id: str
4     study_id: str
5     reference_build: str

TypeScript

typescriptOne Dark Pro

1 export interface SampleManifest {
2   sampleId: string;
3   studyId: string;
4   referenceBuild: string;
5 }

Why interviewers care

Genomics Data Pipeline interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Genomics pipeline framing: from sequencer to clinical report that proves you understand real outages, not happy-path diagrams.

Key Highlights

•120 runs/day
•Sequencer → LIMS → Orchestrator
•Instrument uploads demultiplexed FASTQ to a landing zone; LIMS binds sam

1	@dataclass(frozen=True)
2	class SampleManifest:
3	sample_id: str
4	study_id: str
5	reference_build: str

1	export interface SampleManifest {
2	sampleId: string;
3	studyId: string;
4	referenceBuild: string;
5	}