RAGCell: Retrieval-Augmented Generation as Supervision for Versatile Single-cell Analysis
Abstract
Single-cell foundation models (scFMs) are transforming computational biology by enabling generalizable, task-agnostic representations for versatile single-cell analysis. Despite their progress in facilitating rapid deployment for downstream tasks, off-the-shelf scFMs still have some overlooked concerns: (I) (Pretraining Cost.) Pretrain-based scFMs necessitate pretraining on a vast volume of cells, rendering it draining resources in applications. (II) (Heterogeneous Gap.) Large Language Models (LLM)-based scFMs ignore the tremendous heterogeneous gap between LLM textual and raw cellular spaces, leading to insufficient capability when facing downstream tasks. To this end, we introduce RAGCell, a versatile single-cell analysis framework that achieves a double-win in both cost-effectiveness and high performance. The success of RAGCell lies in two key aspects: Leveraging LLMs to construct cell-level and feature-level knowledge databases, which serve as supervision signals for training the cell model and significantly reduce the training cost (>pretrain-based scFMs). Aligning cell representations with text embeddings from the bi-level knowledge databases, enabling knowledge transfer from textual spaces to cellular spaces and effectively mitigating the heterogeneous gap (>LLM-based scFMs). Through extensive experiments on six downstream single-cell analysis tasks, we demonstrate that RAGCell achieves outstanding performance compared to state-of-the-art scFMs while operating at less than ∼1/10 the cost of pretrain-based scFMs.
Related
- Hippocampo-neocortical interaction as compressive retrieval-augmented generation
- GEO-BERT: A Retrieval augmented LLM Framework for Intelligent Dataset Retrieval from the Gene Expression Omnibus
- QProtFuse: Query-Guided Layer Fusion for Retrieval-Augmented Protein Function Prediction
- Retrieval-Augmented Parkinson's Disease Classification from Resting-State Functional Connectivity
- RA-PLA: Retrieval-augmented graph convolutional networks for protein-ligand binding affinity prediction
BibTeX
@misc{liu2026ragcell,
title = {{RAGCell: Retrieval-Augmented Generation as Supervision for Versatile Single-cell Analysis}},
author = {Tianyu Liu and Fan Zhang and Jiayuan Chen and Kun Wang and Haoxuan Li and Shengju Qian and Zhihong Zhu and Donghao Zhou and Hao Wu and Ziheng Zhang and Zhenxi Lin and Xian Wu and Yefeng Zheng},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.14147},
eprint = {2609.14147},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.14147},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - RAGCell: Retrieval-Augmented Generation as Supervision for Versatile Single-cell Analysis
AU - Tianyu Liu
AU - Fan Zhang
AU - Jiayuan Chen
AU - Kun Wang
AU - Haoxuan Li
AU - Shengju Qian
AU - Zhihong Zhu
AU - Donghao Zhou
AU - Hao Wu
AU - Ziheng Zhang
AU - Zhenxi Lin
AU - Xian Wu
AU - Yefeng Zheng
PY - 2026
DA - 2026/09/12
PB - arXiv
DO - 10.48550/arxiv.2609.14147
UR - https://arxiv.org/abs/2609.14147
AB - Single-cell foundation models (scFMs) are transforming computational biology by enabling generalizable, task-agnostic representations for versatile single-cell analysis. Despite their progress in facilitating rapid deployment for downstream tasks, off-the-shelf scFMs still have some overlooked concerns: (I) (Pretraining Cost.) Pretrain-based scFMs necessitate pretraining on a vast volume of cells, rendering it draining resources in applications. (II) (Heterogeneous Gap.) Large Language Models (LLM)-based scFMs ignore the tremendous heterogeneous gap between LLM textual and raw cellular spaces, leading to insufficient capability when facing downstream tasks. To this end, we introduce RAGCell, a versatile single-cell analysis framework that achieves a double-win in both cost-effectiveness and high performance. The success of RAGCell lies in two key aspects: Leveraging LLMs to construct cell-level and feature-level knowledge databases, which serve as supervision signals for training the cell model and significantly reduce the training cost (>pretrain-based scFMs). Aligning cell representations with text embeddings from the bi-level knowledge databases, enabling knowledge transfer from textual spaces to cellular spaces and effectively mitigating the heterogeneous gap (>LLM-based scFMs). Through extensive experiments on six downstream single-cell analysis tasks, we demonstrate that RAGCell achieves outstanding performance compared to state-of-the-art scFMs while operating at less than ∼1/10 the cost of pretrain-based scFMs.
ER -