Tutorials

Our tutorials demonstrate the practical application of DRVI across different use cases.

• General training and interpretability pipeline
• Transfer learning to new dataset (query to reference mapping)
• Finding rare (un-annotated) cell types with DRVI
• Identifying cell types of DRVI factors
• Identifying biological processes of DRVI factors
• Identifying DRVI factors with LLM tools
• Curating DRVI factor annotations
• Porting a DRVI model (drvi-py < 0.3) to scvi.external.DRVI (scvi-tools)

Installing tutorial dependencies

The tutorials need a few packages on top of the core DRVI stack. These are provided as optional-dependency extras so you only install what a given tutorial needs:

Extra	Install	Covers
tutorials	pip install "drvi-py[tutorials]"	Common packages for the general tutorials (e.g. general pipeline, rare cell types, query-to-reference mapping).
tutorials-cell-types	pip install "drvi-py[tutorials-cell-types]"	Identification of factors — cell types (adds celltypist, networkx).
tutorials-biological-processes	pip install "drvi-py[tutorials-biological-processes]"	Identification of factors — biological processes (adds decoupler, gprofiler-official, gseapy, statsmodels).
tutorials-llm	pip install "drvi-py[tutorials-llm]"	Identification of factors — LLM tools (adds anthropic, cassia, claude-agent-sdk, google-genai, gs2txt, nest-asyncio, openai). Most backends need an API key.
tutorials-all	pip install "drvi-py[tutorials-all]"	Everything for every tutorial, including all identification notebooks and the LLM backends.

Extras can be combined, e.g. pip install "drvi-py[tutorials,tutorials-cell-types]".

Note

The tutorials-llm backends are optional — you only need the one(s) you actually use. Most require an API key, except claude-agent-sdk, which uses a logged-in claude CLI. The “identification of factors — curation” tutorial reads results produced by the earlier notebooks, so it needs no extra packages beyond core DRVI.