This package serves to infer the network structure of the microbiome count data generated by next-generation sequencing (NGS) technology. It also provides functions to generate overdispersed and zero-inflated count data from a Dirichlet Multinomial model with flexible parameter settings. The simulated count data also preserve a pre-specified network structure under the Gaussian graphical model.
install.packages("devtools")
library(devtools)
install_github("shuangj00/HARMONIES", subdir = “pkg”)
example.result = HARMONIES (count.matrix, phenotype, N.mcmc = 100, b = 1, h = 20,
sparsity.cutoff = 0.5, beta.stars = 0.05, n.rep = 20, bayes.fdr = 0.05, seed = 123)
| count.matrix | A count matrix from metagenomic shotgun sequencing or 16SrRNA sequencing technologies. Columns represent the taxa from the same taxonomic level, and rows represent the samples |
| phenotype | A phenotype vector for all the samples. It should be a vector of 0s if only have one phenotype, or a vector of 0 and 1 if have 2 phenotypes |
| N.mcmc | Number of MCMC iterations |
| b | Shape parameter used in the hyper-parameter settings of the variance term |
| h | Scale parameter used in the hyper-parameter settings of the variance term |
| sparsity.cutoff | A threshold between 0-1. Taxa with proportions of zeros larger than the threshold would be dropped for the network inference |
| beta.stars | Stability threshold for selection criteria |
| n.rep | Number of random subsamples to take for network estimation |
| bayes.fdr | Bayesian false discovery rate controlled for zero imputation |
| seed | Random seed |