Gradient Biotech

Study workflow

The product follows a fixed left-to-right flow. You can revisit earlier sections at any time, but downstream results may become stale if inputs change.

Data → Explore → Analyze → Interpret → History

Data — prepare inputs

Upload files, validate ingestion, map metadata, define experimental design, and save contrasts. Nothing in Analyze runs meaningfully until at least one dataset is ready.

Readiness signal: green dot on Datasets when a processed .h5ad exists.

Explore — first look

Inspect QC metrics, sample relationships, embeddings, and spatial layouts before committing to formal analysis. Explore is read-only with respect to the pipeline; it displays outputs from completed runs.

Analyze — compute

Run structured workflows: QC, normalization, clustering, differential expression, enrichment, spatial domains, biomarker discovery. Parameters are saved with every job. Failed runs show actionable errors on the Analyze and History pages.

Interpret — biological meaning

Connect numeric outputs to biology: cluster annotations, pathway enrichment, importable gene sets, disease evidence, and AI-assisted summaries grounded in computed results. Interpret aggregates results from multiple completed runs.

History — provenance

Review active and completed jobs, expand rows for parameters and errors, manage snapshots, and address stale outputs. Deleting runs is blocked when a snapshot references them or while a job is still running.

Typical single-cell path

StepSectionAction
1DataUpload .h5ad or convert 10x/CSV
2DataSave metadata and contrasts
3AnalyzeQC → Normalization → Clustering
4ExploreReview UMAP and QC charts
5AnalyzeDE → Enrichment
6InterpretReview enrichment and gene sets
7HistoryReview run records and create snapshot

See section-specific guides: Data, Explore, Analyze, Interpret, History.