Clarity: An R package for Comparing SimiLARITY Matrices

Overview

See the paper titled CLARITY: comparing heterogeneous data using dissimilarity for more details.

Clarity is an R package for comparing matrices in terms of their structure. If the matrices represent dissimilarities between objects, then the matrix can be decomposed into a structure and a relationship matrix, in the manner of Non-negative Matrix Factorization. Each object is represented in terms of:

• the Structure: the topology of a graph (mixture of trees) that generates the objects in a latent space
• a Relationship matrix represents the branch lengths between the latent structures.

Structural comparison allows for the the structures of two matrices to be compared whilst making minimal assumptions regarding the relationship matrix.

Clarity offers tools to rapidly perform the graph comparison, to identify nodes that are outlying and to identify elements of the matrices that are unusual with respect to one another.

Installation

We provide two packages:

• Clarity, which is an inference framework, and
• ClaritySim, which is a simulation tool for this framework.

Both are installed with the following command:

remotes::install_github("danjlawson/CLARITY/Clarity")

Usage

You should find usage straightforward if you follow the examples in the Clarity package:

help(Clarity)

will point you towards the inline help. The simple usage example is:

## Run the model on your similarity matrix, here the example "dataraw"
scan=Clarity_Scan(dataraw)
## Predict a new similarity matrix, here the example "datamix"
predmix=Clarity_Predict(datamix,scan)
## Plot using the plot.Clarity method:
plot(predmix)

Additional information

The file makeclarity.R provides the code that generated the example datasets.

Citation

Currently cite the package with citation("Clarity").

License and Authors

Clarity and ClaritySim are released under the GPLv3 license.

Clarity is written by Daniel Lawson ([email protected]).