DATE2023.08.01 #Press Releases

Deep learning helps identify cell types based on their gene expression

A new method called scDeepInsight can identify cell types from their gene expression data at 87.5% accuracy. 

Aug 01, 2023  

A Japanese team, including researchers from the University of Tokyo, introduced a new method to identify cell types from single-cell RNA sequencing data. The new method, scDeepInsight, uses a deep learning algorithm and is 7% more accurate than the other cell-type identification methods. It helps better understand complex biological systems and various diseases. 

Single-cell RNA sequencing (scRNA-seq) quantifies the mRNA content of individual cells and their gene expression. It allows us to understand tissues and the types of cells within them. Due to the large-scale data produced by scRNA-seq, manual identification of the cell types becomes challenging and error-prone. That’s why Tatsuhiko Tsunoda and his team relied on a machine learning algorithm called deep learning to discern abstract patterns hidden in the data.  

The team used supervised learning, which means they used reference datasets to train the algorithm to annotate cell types. The algorithm learns the expression patterns of multiple genes in different cells with the team’s new methodology, and based on that, it can identify cell types from the gene expression data alone. scDeepInsight can also identify marker genes that determine the properties of different cell types. It can even recognize small differences between similar cell types. The method can identify rare and unknown cell types that appear in diseases, which will help develop treatments for them. 

Figure. An overview of scDeepInsight method. 

For more details, please read the article:  

Shangru Jia, Artem Lysenko, Keith A Boroevich, Alok Sharma, Tatsuhiko Tsunoda. 2023. scDeepInsight: a supervised cell-type identification method for scRNA-seq data with deep learning. Briefings in Bioinformatics. DOI: 10.1093/bib/bbad266