The provided text is an excerpt from a scientific paper introducing Nicheformer, a novel transformer-based foundation model designed for analyzing single-cell and spatial omics data. This model integrates dissociated single-cell and targeted spatial transcriptomics data from both human and mouse, creating the SpatialCorpus-110M, which contains over 110 million cells for pretraining. Nicheformer aims to capture local dependencies in the cellular microenvironment, excelling in downstream tasks such as spatial composition prediction and spatial label prediction, which traditional models struggle with due to their lack of spatial context awareness. The authors demonstrate that Nicheformer significantly outperforms existing foundation models and traditional embedding methods in these tasks, showcasing its ability to transfer spatial context information to dissociated single-cell data. Overall, the work establishes Nicheformer as a significant step toward developing multiscale machine-learning models for comprehensive spatial single-cell analysis.
Bioinforere
Bioinforere
The provided text is an excerpt from a scientific paper introducing Nicheformer, a novel transformer-based foundation model designed for analyzing single-cell and spatial omics data. This model integrates dissociated single-cell and targeted spatial transcriptomics data from both human and mouse, creating the SpatialCorpus-110M, which contains over 110 million cells for pretraining. Nicheformer aims to capture local dependencies in the cellular microenvironment, excelling in downstream tasks such as spatial composition prediction and spatial label prediction, which traditional models struggle with due to their lack of spatial context awareness. The authors demonstrate that Nicheformer significantly outperforms existing foundation models and traditional embedding methods in these tasks, showcasing its ability to transfer spatial context information to dissociated single-cell data. Overall, the work establishes Nicheformer as a significant step toward developing multiscale machine-learning models for comprehensive spatial single-cell analysis.
1 month ago | [YT] | 0
View 0 replies