Getting started

Quick start.

A first scRNA-seq analysis, from an .h5ad file to a clustered UMAP you can tune. Three ways in — the Analysis Wizard, a template, or the Node Library — all end on the same canvas.

Three ways to start

Analysis Wizard

Pick your data type → pick what you want to do → review & build. A runnable pipeline lands on the canvas; click its Load node and choose your file.

Templates (33)

Whole pipelines, pre-wired with sensible parameters, for every modality. The loader path is empty until you point it at your data.

Node Library

Build by hand. The library is organised by modality (5 views) in pipeline order, and a suggestions panel proposes what usually comes next.

Core concepts

Node	One analysis step with its own parameters, status and result. Drag it from the Node Library onto the canvas.
Edge	Data flowing from one node’s output handle to the next node’s input. Some nodes take two inputs (Merge Data, Cell Mapper).
Workflow	A connected graph that runs in dependency order. One project can hold several workflows, each in its own tab.
Workspace	A modal on a finished node where you tune thresholds or a figure without re-running the analysis: QC Workspace, Visualization Workspace, Embedding Workspace, Obs Inspector, Spatial Viewer.

Build your first workflow by hand

When a new workflow tab opens, the Node Library asks “What are you analyzing?”. Pick Single-cell RNA-seq: the library then shows the scRNA-seq nodes in stages, loader first. The search box always searches every node, and All nodes switches to the classic category layout.

01
Load data
Drag Load H5AD from the Load Data stage onto the canvas. In the Properties panel on the right, pick your .h5ad file. Other scRNA-seq inputs: 10x H5, 10x MTX, CSV/TSV, Loom, plain text, velocyto .loom and STARsolo Velocyto output.
02
Quality Control
Add Quality Control and connect it to the loader. Run it once in label mode, then open the QC Workspace from the node card: it recommends thresholds from the distributions (MAD-based), and moving a slider updates the plots and the cell count live, without re-running. Apply when the thresholds look right.
03
Preprocess
Chain Normalize → Log1p → Highly Variable Genes → Scale → PCA. Defaults are the standard scanpy workflow; Scale keeps only the highly variable genes in X and preserves the full gene set in adata.raw for plots and DEG.
04
Embed and cluster
Add Neighbors after PCA, then connect it to both UMAP and Clustering. Leiden with resolution 0.5–1.0 is a normal starting point (higher means more, smaller clusters); K-Means with an exact n_clusters is there when you need a fixed k.
05
Run
Press Run in the toolbar. Nodes execute in dependency order and turn green as they finish; Stop aborts after the current node. Before a heavy step starts, the app predicts its memory and either proceeds, warns, or refuses with a plain-language reason — see Memory & large data.
06
Look at the result
Add Embedding Plot (or Cluster Visualization) after Clustering, choose the cluster column, and run it. Open its workspace from the card to colour by a gene, rename groups, move the legend, or export — without re-running anything upstream. Then ask the AI Copilot what the clusters look like.

Change one thing, re-run only that

Editing a parameter marks that node and everything downstream as stale. The next Run recomputes only the stale part and reuses every upstream result, so iterating on clustering resolution costs seconds, not a full pipeline.

The screen

Left	Node Library — modality picker, staged pipeline view, search, and the suggestions panel for the selected node.
Centre	The canvas, inside a tab bar. Workflow tabs and Figure Canvas tabs; drag a tab to the side to split the view.
Right	Properties panel for the selected node — parameters, the Setup card when a resource is missing, and the button that opens the node’s workspace.
Bottom	Run Log, AI Chat and Methods tabs. Action plans from the AI arrive in the chat tab as Apply cards.
Toolbar	Run and Stop, Templates and the Analysis Wizard, Report, the project bar with Save / Open / Export, the memory chip and profile menu, the resource status chip, and the hub icon that opens “Connect an AI assistant”.

Where to go next

PBMC tutorial — the same workflow, step by step, with marker genes and annotation.
Canvas, tabs & branches — running only what changed, notes, branch labels, shortcuts.
Visualization Workspace — the five sections every figure shares.
Node reference — every parameter and default.