henhaobio
Sign in

bioRxivBioinformaticsPreprint

Cell2Sentence: Teaching Large Language Models the Language of Biology

Daniel M. LevineiD, Syed Asad RizviiD, Sacha LévyiD, Nazreen PallikkavaliyaveetiliD, David Zhang, Xingyu CheniD, Sina Ghadermarzi, Ruiming WuiD, Zihe ZhengiD, Ivan Vrkic, Anna Zhong, Daphne Raskin, Insu HaniD, Antonio Henrique de Oliveira Fonseca, Josue Ortega CaroiD, Amin KarbasiiD, Rahul M. DhodapkariD, David van DijkiD

PDFDOI
Structured summary available · Jump to summary ↓

Abstract

We introduce Cell2Sentence (C2S), a novel method to directly adapt large language models to a biological context, specifically single-cell transcriptomics. By transforming gene expression data into "cell sentences," C2S bridges the gap between natural language processing and biology. We demonstrate cell sentences enable the fine-tuning of language models for diverse tasks in biology, including cell generation, complex cell-type annotation, and direct data-driven text generation. Our experiments reveal that GPT-2, when fine-tuned with C2S, can generate biologically valid cells based on cell type inputs, and accurately predict cell types from cell sentences. This illustrates that language models, through C2S fine-tuning, can acquire a significant understanding of single-cell biology while maintaining robust text generation capabilities. C2S offers a flexible, accessible framework to integrate natural language processing with transcriptomics, utilizing existing models and libraries for a wide range of biological applications.

Abstract summary

Claude Opus 5.5 · xhigh effort · 2 Nov 2026 · from the abstract · CC BY-NC-ND 4.0

Cell2Sentence converts single-cell gene expression profiles into text-like cell sentences so standard large language models like GPT-2 can be fine-tuned for cell generation and cell-type annotation.

Adapting pretrained natural language processing models directly to single-cell transcriptomics requires representing numerical gene expression profiles in a format compatible with text-based architectures. Cell2Sentence (C2S) converts single-cell expression profiles into cell sentences, allowing existing language models and software libraries to be fine-tuned for biological tasks.

Key results

  1. ClaimedFine-tuning GPT-2 on Cell2Sentence representations enables both conditional generation of biologically valid cells from cell-type prompts and accurate cell-type prediction from cell sentences.

    Reported for GPT-2 fine-tuned with C2S across cell generation, complex cell-type annotation, and data-driven text generation tasks; no numerical accuracy metrics or dataset sizes are given in the abstract. · Abstract

Methods, data and reproducibility

Cell2Sentence (C2S) transforms single-cell gene expression data into sequential 'cell sentences' and fine-tunes pretrained large language models, specifically GPT-2, for cell generation, cell-type annotation, and data-driven text generation.

  • Single-cell transcriptomics gene expression datasets converted into cell sentences
Methods & tools
Cell2Sentence (C2S), GPT-2
Reproducibility gaps
Abstract only; specific single-cell datasets, evaluation metrics, and code repository links are not provided in the abstract.
Limitations (1)
  • The abstract does not state how gene expression values are converted into cell sentences or provide quantitative benchmark results, dataset names, or baseline comparisons. · noted in this summary

Read this if you want to fine-tune standard natural language large language models on single-cell RNA-seq data for cell-type annotation or synthetic cell generation.

Citations
63

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISXiao '25Chen '24Choi '26Rizvi '25Li '24Fang '24
73% semantic matchBioinformatics
Chuxi Xiao, Yuang Ding, Haiyang Bian+3 more·bioRxiv·Oct 22, 2025· 1 citations