Cell2Sentence

⚠️ Status: partial | Version: v1.0

Overview

Converts cells to text sentences for LLM fine-tuning, 768-dim LLM embeddings

!!! tip “When to choose Cell2Sentence”

User wants to leverage general-purpose LLMs, convert cells to text, or use LLM fine-tuning workflows

Specifications

Property

Value

Model

Cell2Sentence

Version

v1.0

Tasks

embed

Modalities

RNA

Species

human

Gene IDs

symbol

Embedding Dim

768

GPU Required

Yes

Min VRAM

16 GB

Recommended VRAM

32 GB

CPU Fallback

No

Adapter Status

⚠️ partial

Quick Start

import omicverse as ov

# 1. Check model spec
info = ov.fm.describe_model("cell2sentence")

# 2. Profile your data
profile = ov.fm.profile_data("your_data.h5ad")

# 3. Validate compatibility
check = ov.fm.preprocess_validate("your_data.h5ad", "cell2sentence", "embed")

# 4. Run inference
result = ov.fm.run(
    task="embed",
    model_name="cell2sentence",
    adata_path="your_data.h5ad",
    output_path="output_cell2sentence.h5ad",
    device="auto",
)

# 5. Interpret results
metrics = ov.fm.interpret_results("output_cell2sentence.h5ad", task="embed")

Input Requirements

Requirement

Detail

Gene ID scheme

symbol

Preprocessing

Requires fine-tuning on reference data. Gene expression is converted to ranked gene sentences.

Data format

AnnData (.h5ad)

Output Keys

After running ov.fm.run(), results are stored in the AnnData object:

Key

Location

Description

X_cell2sentence

adata.obsm

Cell embeddings (768-dim)

 
import scanpy as sc

adata = sc.read_h5ad("output_cell2sentence.h5ad")
embeddings = adata.obsm["X_cell2sentence"]  # shape: (n_cells, 768)

# Downstream analysis
sc.pp.neighbors(adata, use_rep="X_cell2sentence")
sc.tl.umap(adata)
sc.tl.leiden(adata, resolution=0.5)
sc.pl.umap(adata, color=["leiden"])

Resources

Hands-On Tutorial

For a step-by-step walkthrough with code, see the Cell2Sentence Tutorial Notebook.