Reusability report: Exploring the transferability of self-supervised learning models from single-cell to spatial transcriptomics

doi:10.1038/s42256-025-01097-5

	Reusability report: Exploring the transferability of self-supervised learning models from single-cell to spatial transcriptomics
	Han, Chuangyi 1,2,3; Lin, Senlin 1,2,3,4; Wang, Zhikang 1,2,3,5,6; Cui, Yan 1,2,3; Zou, Qi 1,2,3,7; Yuan, Zhiyuan 1,2,3
	2025-08-21
发表期刊	NATURE MACHINE INTELLIGENCE
页码	23
摘要	Self-supervised learning (SSL) has emerged as a powerful approach for learning meaningful representations from large-scale unlabelled datasets in single-cell genomics. Richter et al. evaluated SSL pretext tasks on modelling single-cell RNA sequencing (scRNA-seq) data, demonstrating the effective use of SSL models. However, the transferability of these pretrained SSL models to the spatial transcriptomics domain remains unexplored. Here we assess the performance of three SSL models (random mask, gene programme mask and Barlow Twins) pretrained on scRNA-seq data with spatial transcriptomics datasets, focusing on cell-type prediction and spatial clustering. Our experiments demonstrate that the SSL model with random mask strategy exhibits the best overall performance among evaluated SSL models. Moreover, the models trained from scratch on spatial transcriptomics data outperform the fine-tuned SSL models on cell-type prediction, highlighting a domain gap between scRNA-seq and spatial transcriptomics data whose underlying causes remain an open question. Through expanded analyses of multiple imputation methods and data degradation scenarios, we demonstrate that gene imputation would degrade SSL model performance on cell-type prediction, an effect that is exacerbated by increasing data sparsity. Finally, integrating zero-shot random mask embeddings into chosen spatial clustering methods significantly enhanced their accuracy. Overall, our findings provide valuable insights into the limitations and potential of transferring SSL models to spatial transcriptomics and offer practical guidance for researchers leveraging pretrained models for spatial transcriptomics data analysis.
DOI	10.1038/s42256-025-01097-5
收录类别	SCI
语种	英语
资助项目	Computational Biology Program[25JS2850200] ; Science and Technology Commission of Shanghai Municipality (STCSM), National Natural Science Foundation of China[62303119] ; Science and Technology Commission of Shanghai Municipality (STCSM), National Natural Science Foundation of China[32470706] ; Shanghai Science and Technology Development Funds[23YF1403000] ; Fund of Fudan University[24FCA10]
WOS研究方向	Computer Science
WOS类目	Computer Science, Artificial Intelligence ; Computer Science, Interdisciplinary Applications
WOS记录号	WOS:001555869000001
出版者	NATURE PORTFOLIO
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/41791
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Yuan, Zhiyuan
作者单位	1.Fudan Univ, Shanghai Pudong Hosp, Inst Sci & Technol Brain Inspired Intelligence, Ctr Med Res & Innovat,Pudong Med Ctr, Shanghai, Peoples R China 2.Fudan Univ, MOE Frontiers Ctr Brain Sci, MOE Key Lab Computat Neurosci & Brain Inspired Int, Shanghai, Peoples R China 3.Fudan Univ, Ctr Integrat Spatial Omics Res, Shanghai, Peoples R China 4.Chinese Acad Sci, Inst Comp Technol, Beijing, Peoples R China 5.Monash Univ, Biomed Discovery Inst, Melbourne, VIC, Australia 6.Monash Univ, Monash Data Futures Inst, Dept Biochem & Mol Biol, Melbourne, Vic, Australia 7.Shandong Univ, Ctr Intelligent Med, Sch Control Sci & Engn, Jinan, Peoples R China
推荐引用方式 GB/T 7714	Han, Chuangyi,Lin, Senlin,Wang, Zhikang,et al. Reusability report: Exploring the transferability of self-supervised learning models from single-cell to spatial transcriptomics[J]. NATURE MACHINE INTELLIGENCE,2025:23.
APA	Han, Chuangyi,Lin, Senlin,Wang, Zhikang,Cui, Yan,Zou, Qi,&Yuan, Zhiyuan.(2025).Reusability report: Exploring the transferability of self-supervised learning models from single-cell to spatial transcriptomics.NATURE MACHINE INTELLIGENCE,23.
MLA	Han, Chuangyi,et al."Reusability report: Exploring the transferability of self-supervised learning models from single-cell to spatial transcriptomics".NATURE MACHINE INTELLIGENCE (2025):23.