TripletCell: a deep metric learning framework for accurate annotation of celltypes at the single-cell level.

Liu Y, Wei G, Li C, Shen LC, Gasser RB, Song J, Chen D*,Yu DJ
Brief Bioinform. 2023 May 19;24(3):bbad132. doi: 10.1093/bib/bbad132.

Single-cell RNA sequencing (scRNA-seq) has significantly accelerated theexperimental characterization of distinct cell lineages and types in complextissues and organisms. Cell-type annotation is of great importance in most ofthe scRNA-seq analysis pipelines. However, manual cell-type annotation heavilyrelies on the quality of scRNA-seq data and marker genes, and therefore can belaborious and time-consuming. Furthermore, the heterogeneity of scRNA-seqdatasets poses another challenge for accurate cell-type annotation, such as thebatch effect induced by different scRNA-seq protocols and samples. To overcomethese limitations, here we propose a novel pipeline, termed TripletCell, forcross-species, cross-protocol and cross-sample cell-type annotation. Wedeveloped a cell embedding and dimension-reduction module for the featureextraction (FE) in TripletCell, namely TripletCell-FE, to leverage the deepmetric learning-based algorithm for the relationships between the reference geneexpression matrix and the query cells. Our experimental studies on 21 datasets(covering nine scRNA-seq protocols, two species and three tissues) demonstratethat TripletCell outperformed state-of-the-art approaches for cell-typeannotation. More importantly, regardless of protocols or species, TripletCellcan deliver outstanding and robust performance in annotating different types ofcells. TripletCell is freely available at We believe that TripletCell is areliable computational tool for accurately annotating various cell types usingscRNA-seq data and will be instrumental in assisting the generation of novelbiological hypotheses in cell biology.

Cover image


School of Life Sciences, Nanjing University
Nanjing 210023, China

Back to top