Abstract
Single-cell RNA-seq datasets are often first analyzed independently without harnessing model fits from previous studies, and are then contextualized with public data sets, requiring time-consuming data wrangling. We address these issues with sfaira, a single-cell data zoo for public data sets paired with a model zoo for executable pre-trained models. The data zoo is designed to facilitate contribution of data sets using ontologies for metadata. We propose an adaption of cross-entropy loss for cell type classification tailored to datasets annotated at different levels of coarseness. We demonstrate the utility of sfaira by training models across anatomic data partitions on 8 million cells.
Highlights
Many single-cell data sets are currently published in various databases in different formats, such as custom formats on GEO, manuscript supplements with tables of cell type annotations, or streamlined formats on Human Cell Atlas servers
We identify two core issues with the current state of data and model re-use in single-cell genomics
The sfaira model zoo is designed to be model agnostic and to be as a unified front-end for serving and receiving models, thereby enabling transfer of models from developers to users. In addition to these practical advantages of a data and model zoo, we address the issue of interpretability and generalizability of models
Summary
Many single-cell data sets are currently published in various databases in different formats, such as custom formats on GEO, manuscript supplements with tables of cell type annotations, or streamlined formats on Human Cell Atlas servers. In smaller data sets, rare cell states can often only be properly analyzed after integration with larger reference atlas data sets. This integration is time-intensive and requires a prior analysis of the reference data set. Data processing and cell type annotation are repeated elements of these pipelines that are time-intensive for analysts because of the complexity of the pipelines used. Data processing and cell type annotation are repeated elements of these pipelines that are time-intensive for analysts because of the complexity of the pipelines used4 Both computing an embedding and clustering require basic preprocessing, such as scaling, logtransformation, and highly variable feature selection.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
