Performing Comprehensive Overlap Assessments

The implementation of a statistical framework for performing overlap assessments on lists comprising sets of strings (such as lists of gene sets) described in Stoica (2023) < https://ora.ox.ac.uk/objects/uuid:b0847284-a02f-47ee-88e3-a3c4e0cdb8b1>. It can assess overlaps of pairs of sets of strings selected either from the same universe or from different universes, and overlaps of triplets of sets of strings selected from the same universe. Designed for single-cell RNA-sequencing data analysis applications, but suitable for other purposes as well.

LISTO

LISTO is a tool for performing comprehensive overlap assessments on lists comprising sets of strings, such as lists of gene sets. It can assess:

• Overlaps of pairs of sets of strings selected from the same universe.
• Overlaps of pairs of sets of strings selected from different universes.
• Overlaps of triplets of sets of strings selected from the same universe.

While LISTO has been developed with scRNA-seq data analysis in mind, the methodology is fully applicable for the same problem arising in any other setting. Thus, the implementation of LISTO uses general R objects (data frames, character vectors), rather than scRNA-seq-specific objects.

Installation

To install the version of LISTO currently available on CRAN, run the following R code :

install.packages("LISTO")

Alternatively, you can install the most recent development version, using this code:

pak::pak("andrei-stoica26/LISTO")

Currently, both versions are the same (0.8.1).

Description and usage

This section will elaborate on the functionality and usage of LISTO. It discusses first the overlaps of individual elements, then the details of how the lists of elements must be provided as input.

Items

Each item taking part in an individual overlap assessed by LISTO is a set of strings. Each overlap assessment of sets of strings answers the question of whether the sets intersect each other to a statistically significant extent.

Lists

The runLISTO function runs the entire LISTO pipeline. It requires two lists as input. Each list can store two types of elements:

• Character vectors.
• Data frames with a numeric column specified by the numCol parameter.

A third list, containing the same type of elements, can be optionally provided.

Extracting items from lists

Items to be used in the overlap assessments are extracted from the input lists as follows:

• Character vectors: They are used as such.

• Data frames: The rownames of the data frame are selected, and overlaps are calculated based on cutoffs determined by the distinct values in the column specified by numCol. The median of the resulting p-values is taken to be the p-value of the corresponding overlap.