CoNI is an R package for unsupervised integration of two numerical omics datasets, designed to be practical and easy to use. It employs a correlation-based approach, specifically, partial correlation analysis, to uncover how features in one dataset might influence relationships between features in the other. By calculating partial correlations, CoNI can identify putative confounding variables in one omics layer (the “linker” dataset) that potentially explain or drive the observed correlations among paired features in another layer (the “vertex” dataset). In simpler terms, CoNI finds features in one data layer that could be causing or masking correlations in the other layer, helping researchers tease out hidden associations without any supervised guidance.
After identifying these cross-omics relationships, CoNI generates an interpretable network that integrates the two data layers. This integrated network can be represented in three complementary forms for analysis and visualization: as a weighted undirected graph (with features as nodes and confounding links as weighted edges), as a bipartite graph (where nodes are separated into two types corresponding to each omics layer and connections show their interactions), or as a hypergraph (in which a single confounding feature may connect multiple target features simultaneously). These versatile network representations make it easier to explore complex interactions across the two omics datasets, helping users to identify key multi-omics relationships, detect confounders, and generate new hypotheses about underlying biological interactions.
CoNI is designed to integrate two numerical omics datasets measured on the same samples (e.g., gene expression + metabolomics). Typical examples:
- Transcriptomics (RNA‑seq) + metabolomics
- Proteomics + metabolomics
- Microbiome abundance + metabolite profiles
Before installing CoNI, a few dependencies are necessary:
dependencies<-c("igraph", "doParallel", "cocor", "tidyverse", "foreach", "ggrepel", "gplots", "gridExtra", "plyr", "ppcor", "tidyr", "Hmisc")
`%notin%`<-Negate(`%in%`)
for(package in dependencies){
if(package %notin% rownames(installed.packages())){
install.packages(package,dependencies = TRUE)
}
}
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("genefilter")CoNI requires Python 3. Ensure that Python 3 is installed and correctly added to your system PATH.
If your system configuration makes installing CoNI natively difficult, a Docker container is an alternative way to get CoNI package running.
Note: Docker Machine has Memory and CPU limits on Mac OS X. To control it, please check instructions either for CLI or for Docker Desktop.
For building a Docker image from the Dockerfile, download the Dockerfile (available in this repo) and run the following command to create it:
docker build -t CoNI.
This will create a CoNI Docker image on your system (Please be patient, as the build may take approximately 30–50 minutes to complete).
You can then run it using the following command:
docker run -d -p 8787:8787 -e PASSWORD=pass --name tigaR -it CoNI
CoNI is distributed under the MIT license. Please read the license before using CoNI, distributed in the LICENSE file.
Publications related to CoNI include:
- Monroy Kuhn, J.M., Miok, V., Lutter, D. (2022), "Correlation guided Network Integration (CoNI), an R package for integrating numerical omics data that allows multiform graph representations to study molecular interaction networks". Bioinformatics Advances, 2(1): 0-42.
Please cite the relevant publications if you use CoNI.