https://doi.org/10.1101/2020.06.15.149559
NECo is a node embedding tool that utilizes multiplex heterogeneous gene phenotype network. NECo allows multi-layer gene and phenotype networks. NECo generates gene embeddings using Gene-Gene, Gene-Phenotype, Phenotype-Gene and Phenotype-Phenotype neibhorhood spaces, using the random walk with restart rankings as node sequences.
Includes functions to generate the walk matrix (transition matrix) for multiplex heterogeneous network.
Includes functions for random walk with restart, and generation of neighborhood for each nodes in the multiplex heterogeneous network.
Includes function that generates Top N neighborhoods for Gene-Gene, Gene-Phenotype, Phenotype-Gene and Phenotype-Phenotype.
Simple R script that provides an example of how to use NECo functions to generate different neigbhorhoods of multiplex heterogeneous network by taking input of individual networks of genes, phenotypes and bipartite network.
Generates NECo embeddings of genes by utilizing different neighborhoods generated by NECoFunctions.R script.
An input file given below is required for main.R to generate Gene-Gene, Gene-Phenotype, Phenotype-Gene and Phenotype-Phenotype neighborhoods for different top N parameters.
type file_name
gene pwy_network
gene expr_network
phenotype phenotype_net1
phenotype phenotype_net2
bipartite gene_phenotype_net
Current implementation utilizes "rds" R file for network layers given below format:
from to weight
gene1 gene2 0.5
For unweighted networks just provide 1 as weight. The related function is read.network.layers in RWFunctions.R file to be modified for other formats.
Run main.R with modified input file and then run GenerateNECoEmb.py to generate embedding file for different neighborhood spaces. Gene embeddings of different embedding spaces then can be concatenated and utilized by a classification algorithm.