README.md

In ChristophRau/wMICA: weighted Maximal Information Component Analysis

wMICA

Weighted Maximal Information Component Analysis

Updated 2/22/2021 to be significantly faster and install more rationally.

Now works for both Unix and Windows-based machines!

Version: 0.9 Author: Christoph Rau

A weighted implementation of Maximal Information Component Analysis

wMICA is an updated form of Maximal Information Component Analysis as originally described here: http://www.ncbi.nlm.nih.gov/pubmed/23487572

Usage: devtools::install_github("ChristophRau/wMICA")

MICA is used for the analysis of large interconnected networks and the identification of modules of similarly-acting nodes within the larger network. It was designed specifically to work with gene expression datasets (RNA microarrays, RNAseq, etc), but is expandable to any relational network. wMICA avoids two common pitfalls by using the Maximal Information instead of Pearson correlation to calulate node relatedness, which preserves non-linear relationships in the data, as well as a interaction component modeling process to allow each node to proportionally have membership in all modules.

In addition to transforming the algorithm from a series of disconnected scripts to a single package with a single master function (wMICA.Run()), the package now optionally implements the following:

Edge weighting: It is possible (and, in fact, highly recommended due to improved Gene Ontology Enrichments and module comprehensibility) to now add the calculated relationship score between two nodes to the module identification step. This has been previously shown in other algorithms to dramatically improve the quality of the results and does the same here.

Module Seeding: If desired, the non-supervised nature of the algorithm may be compromised by allowing the user to 'seed' modules with certain genes to, for instance, pull down a module relating to a gene of interest or tease apart a pathway's interacting partners.

Weighted GO Enrichment harnesses the continuous nature of the module memberships to calculate GO enrichment without the need for a hard threshold. It is implemented as a separate package: https://github.com/ChristophRau/GSAA

KNOWN BUG: if the last (or last several) genes in your dataset do not have strong enough MI scores to pass the threshold for inclusion in the algorithm, then an error will occur. Deleting these columns will fix the problem, and I will patch the problem shortly.

ChristophRau/wMICA documentation built on Feb. 25, 2021, 2:10 p.m.