Gradient Biotech

Interpret

Interpret turns structured analysis outputs into biological context. It reads from completed runs; it does not recompute statistics — every AI-generated section cites only values already present in the underlying run results.

Annotations

Shows cluster- or domain-level summaries with top marker genes, per-cluster annotation scores, and confidence badges from the Cell Type Annotation step. Use AI Summaries (below) or the AICard on a selected cluster in the UMAP explorer for a narrative cell-type suggestion grounded in that cluster's marker genes — always verify it against the marker table before relying on it.

Enrichment

Links to the GO/KEGG enrichment tree, GSEA, and score-summary tabs built from the latest DE or enrichment run. Terms are ranked by significance; click through to gene lists. An Analyze button on the enrichment result page requests an AI narrative over the significant terms.

Gene Sets

Import gene lists from DE, cluster markers, enrichment, coverage DEG, or biomarker panels into a working set for Explore overlays. Saved gene sets persist in your browser for the current study.

Disease Evidence

Maps selected run-derived gene sets to the bundled disease association lookup. Use it to see which known disease categories overlap with DE genes, cluster markers, enrichment gene lists, or biomarker panels. This is research context, not diagnostic evidence.

AI Summaries

Generates narrative summaries grounded in computed results — not generated statistics. Available modes:

ModeGrounded inWhere it's triggered
Cluster annotationMarker genes for a selected clusterAICard in the UMAP cluster panel
DE interpretationTop differential expression genes and directionAnalyze button on the DE results page
Enrichment interpretationSignificant GO/KEGG/GSEA termsAnalyze button on the enrichment results page
Study summary (with methods)All completed run results plus recorded parameters and versionsInterpret page AI Summaries section

Every response is validated against the structured run context before display — the model may not cite a metric, gene, or p-value that isn't in the underlying result payload. If a summary looks unavailable, confirm the study has at least one completed run for that step and that the interpretation service is reachable.

Methods Provenance

Lists parameters, dataset IDs, and run IDs for completed steps. The AI Summaries section's study-summary mode also drafts a methods-section narrative from the same recorded parameters and pipeline versions — review and edit it before using it in a manuscript.

Rules for interpretation

  • Always verify AI-suggested labels and narratives against the underlying marker genes, DE table, or enrichment terms.
  • Check whether results are stale on the History page before using them in downstream interpretation.
  • Enrichment quality depends on gene symbol identifiers matching the reference database (HGNC for human demo data).