On September 16, 2022, the Fifth Annual Summit of Single-Cell Multiomics Research and Clinical Applications was conducted in Shanghai. Professor Liu was invited to give a presentation.
This report reviewed the pipeline of single cell sequencing and challenges in single cell sequencing data analysis. There are multiple algorithms for single cell sequencing, however there is a lack of the real world ground truth for validating the efficacy of those algorithms, which causes the subjective algorithms selection. In this study, the authors pooled 10 types of pre-labeled immune cells for single cell sequencing and obtained the data with the known cell types. And the authors utilized the deep generative models to infer and remove the noise cells deriving from the sequencing process, which improved the consistency of specific genes analysis and the efficacy of clustering. Then, the authors adopted the Deep-LDA based on the latent Dirichlet allocation to couple the clustering and specific genes analysis, which optimized the workflow of single cell sequencing analysis, and the Deep-LDA can be applied for annotation of single cell types.
