%0 Journal Article %1 yang2022scbert %A Yang, Fan %A Wang, Wenchuan %A Wang, Fang %A Fang, Yuan %A Tang, Duyu %A Huang, Junzhou %A Lu, Hui %A Yao, Jianhua %D 2022 %I Cold Spring Harbor Laboratory %J Nature Machine Intelligence %K dmir-readinggroup machine-learning single_cell %P 852-866 %R 10.1038/s42256-022-00534-z %T scBERT as a Large-scale Pretrained Deep Language Model for Cell Type Annotation of Single-cell RNA-seq Data %U https://www.nature.com/articles/s42256-022-00534-z %V 4 %X Annotating cell types on the basis of single-cell RNA-seq data is a prerequisite for research on disease progress and tumour microenvironments. Here we show that existing annotation methods typically suffer from a lack of curated marker gene lists, improper handling of batch effects and difficulty in leveraging the latent gene–gene interaction information, impairing their generalization and robustness. We developed a pretrained deep neural network-based model, single-cell bidirectional encoder representations from transformers (scBERT), to overcome the challenges. Following BERT’s approach to pretraining and fine-tuning, scBERT attains a general understanding of gene–gene interactions by being pretrained on huge amounts of unlabelled scRNA-seq data; it is then transferred to the cell type annotation task of unseen and user-specific scRNA-seq data for supervised fine-tuning. Extensive and rigorous benchmark studies validated the superior performance of scBERT on cell type annotation, novel cell type discovery, robustness to batch effects and model interpretability.

@article{yang2022scbert, abstract = {Annotating cell types on the basis of single-cell RNA-seq data is a prerequisite for research on disease progress and tumour microenvironments. Here we show that existing annotation methods typically suffer from a lack of curated marker gene lists, improper handling of batch effects and difficulty in leveraging the latent gene–gene interaction information, impairing their generalization and robustness. We developed a pretrained deep neural network-based model, single-cell bidirectional encoder representations from transformers (scBERT), to overcome the challenges. Following BERT’s approach to pretraining and fine-tuning, scBERT attains a general understanding of gene–gene interactions by being pretrained on huge amounts of unlabelled scRNA-seq data; it is then transferred to the cell type annotation task of unseen and user-specific scRNA-seq data for supervised fine-tuning. Extensive and rigorous benchmark studies validated the superior performance of scBERT on cell type annotation, novel cell type discovery, robustness to batch effects and model interpretability.}, added-at = {2023-06-05T09:07:13.000+0200}, author = {Yang, Fan and Wang, Wenchuan and Wang, Fang and Fang, Yuan and Tang, Duyu and Huang, Junzhou and Lu, Hui and Yao, Jianhua}, biburl = {https://www.bibsonomy.org/bibtex/2f42324d95994db3273b64f2ca6951ef6/martinr}, doi = {10.1038/s42256-022-00534-z}, elocation-id = {2021.12.05.471261}, eprint = {https://www.biorxiv.org/content/early/2022/08/01/2021.12.05.471261.full.pdf}, interhash = {e321ca7a0b0510e11881fbdd82834aae}, intrahash = {f42324d95994db3273b64f2ca6951ef6}, issn = {2522-5839}, journal = {Nature Machine Intelligence}, keywords = {dmir-readinggroup machine-learning single_cell}, language = {en}, month = {10}, pages = {852-866}, publisher = {Cold Spring Harbor Laboratory}, timestamp = {2023-06-05T09:15:47.000+0200}, title = {scBERT as a Large-scale Pretrained Deep Language Model for Cell Type Annotation of Single-cell RNA-seq Data}, url = {https://www.nature.com/articles/s42256-022-00534-z}, volume = 4, year = 2022 }