A Comprehensive and Versatile Framework for Single-Cell Gene Co-Expression Studies and Cell Type Identification
The estimation of gene co-expression in single-cell RNA sequencing (scRNA-seq) is a critical step in the analysis of scRNA-seq data. The low efficiency of scRNA-seq methodologies makes sensitive computational approaches crucial to accurately infer transcription profiles in a cell population.
COTAN is a statistical and computational method that analyzes the
co-expression of gene pairs at the single-cell level. It employs an innovative
mathematical model that leads to a generalized contingency table framework.
COTAN relies on the zero Unique Molecular Identifier (UMI) counts
distribution instead of focusing on positive counts to evaluate or extract
different scores and information for gene correlation studies and gene or cell
clustering.
COTAN assesses whether gene pairs are correlated or anti-correlated,
providing a new correlation index with an approximate p-value for the
associated test of independence. It also checks whether single genes are
differentially expressed, scoring them with a newly defined Global
Differentiation Index (GDI). COTAN plots and clusters genes according to
their co-expression pattern with other genes to study gene interactions and
identify cell-identity markers.
COTAN v2 introduces a novel feature that uses gene GDI values to assess the
biological uniformity of a cell cluster. This feature allows researchers to
apply an iterative cell clustering pipeline and achieve a finer resolution of
uniform clusters. COTAN shows high sensitivity in extracting information from
small clusters and lowly expressed genes. Furthermore, COTAN leverages its
contingency table framework to directly identify genes that are over-represented
or under-represented in the cluster with respect to the rest of the data-set.
COTAN computes an enrichment score for a given list of marker genes, which can
be used to identify and merge small uniform clusters and to check a final
cluster identification.
From version 2.0.0 new functions and plots to check and clean the data-set were
included along several visualization tools to help users explore and interpret
their data. COTAN has a user-friendly interface that is easy to use and does
not require extensive programming skills. The strength of COTAN is its ability
to help researchers better understand scRNA-seq data. By identifying gene
modules, cell types, and new marker genes, researchers gain insights into the
underlying biology of their samples. This helps disease diagnosis, drug
discovery, and other applications. In summary, COTAN is a powerful and
versatile tool for the analysis of scRNA-seq data, with the potential to
facilitate the discovery of new cell types and biological insights.
Main source of examples for the COTAN v2 is the vignette:
Guided_tutorial_v2{.uri}. There it is illustrated the preparatory cleaning steps,
various analysis results and plots done on the data-set "Mouse Cortex E17.5, <GEO:GSM2861514>"
Further more it is possible to look at some other examples on real data-sets at COTAN paper{.uri} and more extensively at COTAN Datasets analysis{.uri}.
The first link shows how to handle the genes' clustering while the second shows how to use the new cells' clustering functions to obtain uniform clusterizations [Please note: the first link has not been upgraded to the version 2, so, while it is possible to reproduce all the steps there described, they need to be manually adapted to the new interface to be executed]
The production version can be installed via Bioconductor COTAN for Bioconductor 3.18{.uri}, while a stable development version can be found COTAN for Bioconductor 3.19{.uri}.
The latest snapshot can be installed directly as R package using devtools.
The installation was tested on Linux, Windows and Mac but please note that
due to lack of multi-core support under Windows it might run slower.
devtools::install_github("seriph78/COTAN")
From version 2.5.0 COTAN can optionally use the torch library and thus, in
case, use the GPU for some of its calculations, with substantial speed-ups.
However this implies a possibly more complicated installation process: see the
specific help page Installing
torch{.uri}
for some pointers.