Siftome Phase 6A

Analysis Handoff Boundary

Ranked results Import status

Siftome prepares public GEO datasets for downstream bioinformatics review. It does not run raw-data reanalysis or present differential expression results as product output.

Prepared Outputs

4 preparation outputs

  • Dataset triage package Ranked datasets, source links, assay type, organism, sample counts, warnings, recommendation labels, and score rationale.
  • Sample metadata preparation Original GSM titles, source names, characteristics, normalized disease/tissue/sample labels, and proposed case/control hints.
  • Review-ready exports CSV and Markdown exports that preserve source metadata, warnings, grouping hints, and caveats for a downstream analyst.
  • Analysis readiness signals Flags for missing controls, small cohorts, model systems, secondary assays, unclear tissue labels, and metadata that needs manual review.

Explicit Non-goals

5 excluded analysis steps

  • No FASTQ, BAM, or SRA reanalysis Siftome does not download sequencing runs, align reads, quantify expression, or run workflow engines.
  • No raw-analysis execution Siftome has no action that starts FASTQ/BAM/SRA processing, read alignment, quantification, or workflow execution; raw reanalysis decisions must happen outside the product.
  • No expression matrix ingestion yet Processed matrices, count tables, TPM/FPKM values, and differential expression tables are not imported into the product database.
  • No statistical testing Siftome does not run DESeq2, edgeR, limma, survival analysis, enrichment analysis, or clinical-outcome modeling.
  • No biological proof claims Siftome output must not claim mechanism proof, target validation, druggability, diagnostic validity, or clinical actionability.

User-facing Guardrails

4 guardrails
Use preparation language

Pages and exports should say prepared for downstream analysis, needs manual review, or suitable for triage, not analyzed or validated.

Keep source evidence visible

Every handoff artifact must preserve the original GEO accession, source link, raw metadata text, and rule-based warnings.

Separate assay families

Bulk RNA-seq, microarray, secondary RNA assays, biofluid cohorts, and model-system experiments should not be silently mixed.

Require analyst confirmation

Any downstream contrast, grouping, exclusion, or expression-analysis step remains a proposed handoff item until a human analyst confirms it.