Single-cell analysis in Correlation Engine

Single-cell technology enables the analysis of molecular characteristics at the level of individual cells, rather than bulk populations. This approach is used to study a wide range of genomic modalities, including gene expression, protein levels, chromatin accessibility, and DNA methylation-seq. Marker lists or differential analyses from scRNA-seq, scATAC-seq, or scMethylation data can be imported into Correlation Engine, just as with other genomic analyses. The data are processed using directional rank-based statistical algorithms, allowing users to query their signatures against any standard workflow in Correlation Engine. Using Correlation Engine applications for single-cell studies users can:

· Identify cell types and functional identities by querying the Body Atlas and other single studies within the Curated Studies library

· Categorize individual cell types in disease versus normal analysis with the Disease Atlas to determine which cells are likely to contribute to disease pathology

· Identify unique treatment responses in each cell type using the Pharmaco Atlas

· Help elucidate biomolecular mechanisms across cell types with tools such as the Knockdown Atlas, Pathway Enrichment, and Meta-analysis

Single cell data added to the public Curated Studies library are sourced from free-access repositories like the Gene Expression Omnibus (GEO: https://www.ncbi.nlm.nih.gov/geo/) and curated resources from the CZI CELLxGENE project (https://cellxgene.cziscience.com/). To be considered for curation, studies must provide metadata with pre-determined cell identities. Data are re-analyzed from raw counts using Seurat for normalization, variable feature identification, scaling, PCA, UMAP, neighbor discovery, cluster identification, and marker analysis via the Wilcoxon test.

Two comparison designs are considered when curating single marker signatures (referred to as biosets in the application):

· Directed Comparison: A biological test case is provided in the metadata (e.g., disease vs. normal, treatment vs. control, timeline analysis). These cell-type-specific biosets are scored using directional fold changes or other appropriate measures for ranking.

· Atlas-Style Comparison: Comparisons are made between each cell type cluster and all other cell types in the analysis. Here, a ranking statistic is based on the strength of a gene feature p-value, with the directional sign of the fold change applied. This matches the design of biosets in the Body Atlas application. Unlike other Body Atlas biosets, single cell biosets are available for direct inspection and use in Curated Studies.

References

CZ CELLxGENE Discover: A single-cell data platform for scalable exploration, analysis, and modeling of aggregated data. CZI Single-Cell Biology, et al. bioRxiv 2023.10.30; doi: https://doi.org/10.1101/2023.10.30.563174

CELLxGENE: a performant, scalable exploration platform for high-dimensional sparse matrices. CZI Single-Cell Biology, et al. bioRxiv 2021.04.05; doi: https://doi.org/10.1101/2021.04.05.438318

Hao Y, Stuart T, Kowalski MH, Choudhary S, Hoffman P, Hartman A, Srivastava A, Molla G, Madad S, Fernandez-Granda C, Satija R (2023). “Dictionary learning for integrative, multimodal and scalable single-cell analysis.” Nature Biotechnology. doi:10.1038/s41587-023-01767-y, https://doi.org/10.1038/s41587-023-01767-y.