文献阅读 NC：单细胞多组学在免疫肿瘤学中的应用

简要介绍一篇最近发表在Nature Communications上的COMMENT，The use of single-cell multi-omics in immuno-oncology， DOI号：10.1038/s41467-022-30549-4。

1. Integrative analyses of scMulti-omics in immuno-oncology

• Intra-modality integration where the same modality is measured from different cells
• Unmatched inter-modality integration where multiple modalities are measured from different cells, samples, or experiments 【scRNA+scTCR, scRNA+scATAC, protein abundance+gene expression+chromatin accessibility】
• Matched inter-modality integration where multiple modalities are measured from the same cell【CITE-seq (gene expression+protein abundances), scRNA+ST, scRNA+scATAC】

2. Challenges and future prospects

• Batch effect removal is one of the main obstacles in accurate integrative analyses, which needs to retain the true signals and remove differences between samples, conditions, and experiments

Benchmarking methods
(1) Tran, H. T. N. et al. A benchmark of batch-effect correction methods for single-cell RNA sequencing data. Genome Biol. 21, 12 (2020).
(2) Luecken, M. D. et al. Benchmarking atlas-level data integration in single-cell genomics. Nat. Methods 19, 41–50 (2022).

• Multiple computational tools have been developed to integrate scMulti-omics data in a generic style, but not particularly designed or optimized for data analysis in immuno-oncology

For example, the number of features in dimension reduction and the resolution in Louvain clustering can be tailored higher in immuno-oncological scMulti-omics data than those used for normal tissues or cell lines. Also, acknowledged marker genes and signatures (e.g., CD3 and CD4 for CD4+ T cells) can be included in the data analyses to auto-correct the cell clustering result.

• Current methods have limited power to understand the cross-talk between cells and different modalities
• As the data complexity increases (e.g., ten million cells in one dataset), computational efficiency becomes more critical and requires scalability to handle huge amounts of data

Databases, Deep Learning, End-to-End (modularized), Wet-lab validated