November 18, 2022
Keio University School of Medicine
A research group at the Keio University School of Medicine, Department of Physiology, led by Yuji Okano (a fifth-year student at the School of Medicine), Assistant Professor Yoshitaka Kase, and Professor Hideyuki Okano, has developed a model to overcome unresolved technical challenges in single-cell RNA sequencing (scRNA-seq) data analysis, successfully extracting universal features across scRNA-seq datasets.
RNA-seq, which can comprehensively collect and analyze the genetic information of cells, and especially scRNA-seq, which can read the information of each individual cell, have now become indispensable in modern medical research. However, because it is a rapidly developing technology, its analysis methods have several challenges, and it has been questioned whether the results obtained truly reflect the biological essence.
Furthermore, the process of annotation, which infers cell type information from scRNA-seq data, is the most critical step for scRNA-seq data that lacks cellular morphological information. However, conventional annotation methods have been unable to extract the "universal characteristics" of cell types shared across different individuals. This is because they perform relative comparisons between samples within a dataset through differentially expressed gene analysis and infer cell types by arbitrarily selecting interpretable genes from a group of genes with significantly increased expression. Moreover, since annotation is performed prior to various other suggestive data analyses on scRNA-seq data, it has been a problem to what extent the results of scRNA-seq data analysis reflect a generalizable truth.
In this study, we succeeded in developing a metric to evaluate the similarity of cellular characteristics based on network similarity, by visualizing the "universal characteristics" of cell types from scRNA-seq data as a gene regulatory network. Furthermore, by applying this metric to annotation in a systematic manner, we have made it possible to effectively integrate scRNA-seq data derived from different individuals. In fact, when we implemented the model from this research on scRNA-seq data from the human fetal brains of multiple individuals, we found that it enabled annotation that reflects more cross-dataset features compared to conventional methods.
This achievement is expected to yield important and essential results that have been overlooked in previous scRNA-seq studies, such as in disease analysis. Furthermore, its application is anticipated in various fields, as it allows for the confirmation of universal characteristics without arbitrary processing of raw data in research themes that require the integration and analysis of data from samples obtained from multiple individuals or research institutions, such as in the analysis of rare diseases. The results of this research were published in a special issue of Stem Cell Reports , the official journal of the International Society for Stem Cell Research, on November 17, 2022, at 11:00 a.m. (EST).
For the full press release, please see below.