| AI interpretation | Grounded narrative summary (cluster annotation, DE, enrichment, or study/methods text) generated from structured run results; cannot cite values absent from the run payload |
| AnnData | Annotated data matrix format (.h5ad) used internally for all modalities |
| Checkpoint | Intermediate processed .h5ad written between pipeline steps |
| Contrast | A saved comparison between two groups (for example treatment vs control) |
| Coverage DEG | Multi-class biomarker step ranking genes by recurrence across all pairwise class comparisons |
| DE | Differential expression — genes that change between groups or clusters |
| Deconvolution | Estimating cell-type composition fractions from sample- or spot-level expression profiles |
| Disease evidence | Curated gene-disease association lookup applied to run-derived gene sets; research context, not diagnostic evidence |
| Domain | Spatial cluster of spots representing a tissue region |
| Enrichment | Statistical over-representation or ranking of genes against curated pathways (GO, KEGG) |
| Experiment | Database/API name for a study |
| GSEA | Gene set enrichment analysis — ranked whole-DE-table enrichment method, reported alongside ORA |
| HVG | Highly variable gene — selected before dimensionality reduction |
| Integration | Batch-correction step that removes sample/batch-driven structure from a joint PCA/clustering |
| Leiden | Graph-based clustering algorithm used after neighbor graph construction |
| Metadata | Sample or cell annotations merged into AnnData obs |
| Modality | Assay type: single-cell, bulk, spatial, or biomarker |
| Niche | Lightweight spatial candidate region nominated from co-localizing, enriched tissue domains |
| ORA | Over-representation analysis — enrichment method for thresholded gene lists |
| Pipeline run | One execution of an analysis step with saved parameters and results |
| Provenance | Record of how a result was produced (parameters, versions, inputs) |
| Pseudobulk | Aggregating single-cell counts by sample and cell type/cluster, then running replicate-aware DE |
| Pseudotime | Graph-based ordering of cells along an inferred trajectory |
| QC | Quality control — filtering low-quality cells or samples |
| Run | See pipeline run |
| Single-cell CNV | Chromosome-scale copy-number inference from single-cell expression, used for malignant-cell and subclone calls |
| Sketch | Representative-cell downsampling and projection strategy for working with large single-cell datasets |
| Snapshot | Frozen study state referencing specific runs and parameters |
| ssGSEA | Single-sample gene set enrichment — per-sample or per-cell gene-set scores |
| Stale | Result outdated because upstream data or parameters changed |
| Study | Top-level project container in the UI |
| SVG | Spatially variable gene — expression varies across tissue coordinates |
| Trajectory | Ordering of cell states (pseudotime) plus cell-cycle scoring for differentiation/activation analysis |
| UMAP | 2D embedding for visualizing cell similarity |
| WGCNA | Weighted gene co-expression network analysis — modules, hub genes, and eigengene-trait correlations |
| Wilcoxon | Non-parametric test used for rank-based DE in the current pipeline |