Professional onboarding

Your first hour with CelliVerse.

Choose the right workflow, run a minimal analysis, understand how the functions connect and resolve common first-use problems without digging through several separate documents.

This standalone hub brings the most useful adoption material into one interactive page.

Start the workflow Explore CelliVerse Full vignette

What is your starting point?

Choose a starting point to get an immediate recommendation.

Quick start

Minimal reproducible route

For a new dataset, keep the first pass deliberately simple. Discover structure and markers, return the labels to the original object, then annotate with evidence.

clustoCell()

Discover structure + markers

Run joint population discovery and obtain ranked marker evidence for the inferred structure.

addClustoData()

Write labels to metadata

Add the major and/or sub-cluster labels back to your Seurat or SingleCellExperiment object.

typoClust() / typoPrompt()

Annotate with evidence

Use curated MarkerDB or LLM-assisted annotation, or generate a portable prompt for a chatbot of your choice.

Focused analysis comes later: markoClust() and markoCell() are not part of the minimal route. Use them when you already have defined groups or when a specific subset of cells deserves deeper marker analysis.

CelliVerse at a glance

Choose the biological question first.

The core workflows are easier to remember as analytical intentions than as a flat function list.

Open interactive function guide

Discover

Discover populations

clustoCell()

Unknown structure; joint major/sub-clustering plus ranked marker evidence.

Interrogate

Markers for known groups

markoClust()

Predefined groups; identify markers and optionally resolve sub-clusters.

Interrogate

Markers for selected cells

markoCell()

Arbitrary subsets, rare populations, discordant cells or even individual cells.

Annotate

Curated annotation

typoClust(mode = "markerDB")

Positive/negative marker matching against CelliVerse MarkerDB; no LLM or API key required.

Annotate

LLM annotation

typoClust(..."ceLLMarkup") / ceLLMarkup()

Use object markers through typoClust or external marker data through ceLLMarkup.

Annotate

Portable annotation prompt

typoPrompt()

Create an inspectable prompt for your preferred chatbot without calling a model from R.

Common next steps

These helpers usually appear after the main discovery, interrogation or annotation step.

addClustoData()write major/sub-cluster labels to metadata

getDatasetMarkers()select marker-derived features for PCA, UMAP or ML

markerPurity()quantify purity of a candidate marker

featureInspect()locate a feature across marker collections

addTypoData()write cell-type annotations back to metadata

clustoCell_TransferLabel()transfer sketch labels to the full dataset

Interactive function guide

One navigator, three ways to think.

The decision tree, function map and task-to-function lookup are combined here so the same workflow knowledge can be explored by situation, stage or search without repeating content.

Annotation quality rule: automated labels are evidence-supported suggestions, not a replacement for marker biology and context. For broad or mixed parent clusters, compare hierarchical annotation with inheritance disabled; in MarkerDB mode, inherit_score_ratio controls how close runner-up parent candidates must be to constrain sub-clusters.

Troubleshooting

Common problems, fast fixes.

Search by symptom or filter by category. Open a card only when you need the explanation or code.

Natural-language interface

Agent in four steps

The Agent is optional and complementary to standard R scripts. Use it for conversational convenience while keeping formal analyses reproducible.

1Load data

Upload a supported object/file or use a server-side path for large data.

2Choose a model

Select a supported cloud provider or use Ollama/LM Studio for a local workflow.

3Ask clearly

One explicit task at a time is the most reliable pattern, especially for smaller models.

4Review outputs

Inspect generated figures, tables and R objects, then preserve the reproducible analysis trail.

run clustoCell on the objectadd labels to the Seurat objectgenerate a UMAP of the object and color cells by sub-clustersgive me the top 10 ranked markers of C2 and C4annotate clusters C1-C3annotate sub-clusters C1-Sub1 and C3-Sub2