A reliability metric for cell type annotation in single-cell RNA sequencing analysis
Presenting author: Chia-Jung "Charlene" Chang, Department of Biostatistics, Vanderbilt University Medical Center
Co-authored by:
- Chih-Yuan Hsu, Department of Biostatistics, Vanderbilt University Medical Center
- Qi Liu, Department of Biostatistics, Vanderbilt University Medical Center
- Yu Shyr, Department of Biostatistics, Vanderbilt University Medical Center
Abstract:
Single-cell RNA sequencing (scRNAseq) provides an unprecedent opportunity to study heterogeneity and dynamics inherent in cellular compositions. An essential step in scRNAseq analysis is to annotate cell types automatically. Significant effort has been devoted to develop annotation tools, with the current repertoire including over one hundred specialized packages and approaches. However, there is lack of standardized metrics to evaluate their accuracy and reliability consistently. To bridge this gap, we develop a reliability metric for cell type annotation by comparing module score distributions between query samples and their references. Our preliminary evaluations on PBMC datasets across different platforms have demonstrated its value for improving the interpretation of cell type annotation results.