In robertamezquita/Inkwell: Containers for Curated Cell Type Signatures

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

The Form of Signatures

See concepts from the Hancock package for more from @kevinrue. In brief, signatures can be define qualitatively or quantitatively. Additionally, markers may include both positive and negative markers. Furthermore, these markers may be absolute or relative.

Defining Signatures

This package seeks to define signatures which can be used downstream to apply them onto novel (single-cell RNA-seq) datasets. This can be done a number of ways, explored below.

De novo Signatures

Signatures can be defined ad hoc or de novo based on internal or external (published) data. The approach will vary depending on the type of data used.

• For bulk RNA-seq experiments, it is preferable to define signatures from sorted populations, and thus the cell surface/protein markers/biology of the experiment be included in the signature definition. One concern is the applicability of bulk RNA-seq onto single-cell data. This can be remediated somewhat by using a pseudobulking strategy to aggregate a cluster of cells from scRNA-seq, however, it is unclear to what degree this ameloriates the single-cell effect.
• For single-cell RNA-seq experiments, the ideal would be the concurrent use of well known cell surface markers via platforms such as Ab-seq/CITE-seq, clustering cells primarily on this basis to define RNA markers

Predefined Signatures

This depends on having an externally validated gene expression signatures for use. This can come from databases/websites/publications.

• Manual definitions from publications or other non-programmatically accessible resources can be handwritten into a compliant geneset format
• Select web resources can be programmatically accessed to generate compliant geneset format signatures

Package Objectives

The packages that are currently envisioned to be a part of this cell type signature project is:

• GSEABase (or other) to store signatures
• Inkwell for methods to define signatures
• Hancock for methods to assign signatures

Resources

Potential resources include:

• Haemosphere - includes the Haemopedia Human/Mouse atlases and Immgen for mice and more; these are bulk RNA-seq experiments
• DICE - has select cell types available with protein markers and top unique genes from their analyses
• SHoGOIN - comprehensive repository of human annotated resources

robertamezquita/Inkwell documentation built on May 17, 2019, 10:13 a.m.