README.md

In hwang655/HUG: Dependent Graphical Model Estimation

HUG

author: Jianyu Liu, Haodong Wang

Overview

HUG is a data analysis package for estimating Poisson and Hurdle graphical models. The proposed graphical model is designed to be appropriately modeling sample dependence and possible zero-inflation in the data. The sample dependence means there is dependence across the units of analysis (e.g. individuals, cells). The zero-inflation refers to the case when the number of zeros is so large that the data do not readily fit standard distributions. It contains two functions. - MPoisGraph, which fit a dependent Poisson graphical model via nodewise random effect Poisson regression. - MHurdGraph, which Fit a dependent Hurdle graphical model via nodewise random effect Hurdle regression.

Prerequisite

HUG package requires R version 3.6.0 or higher. It requires the packages Matrix, foreach, doParallel, qlcMatrix, fields, mvtnorm, glmnet, huge, igraph, fields, doSNOW

Installation

The HUG package is currently available using devtools

# install.packages('devtools')
devtools::install_github("hwang655/HUG")

Example

An example use: ``` library(HUG)

mb.dep.pois = MPoisGraph(Y, Y.p, Omega = NULL, B.ini, lambda.min.ratio=5e-3) mb.dep.hurd = MHurdGraph(Y, Y.p, Omega = NULL, B.ini, lambda.min.ratio=5e-3) where *Yis the response of nodewise regressions, e.g. scRNA-seq data matrix with rows for cells and columns for genes, *Y.pis the predictors of nodewise regressions, which has the same dimension as Y.r. *B.iniis a reasonable initial coefficient matrix for the nodewise Hurdle regression *lambda.min.ratiois ratio between the minimum and the maximum of the lambda sequence *Omega``` is the precision matrix of the sample dependence model

The output is a list with five elements

• lambda is the the lambda sequence used in nodewise regressions
• graphs is a sequence of estimated graphs
• coef.opt is a pxp matrix of EBIC-selected coefficient estimates of all nodewise regressions
• coef.aic is a pxp matrix of AIC-selected coefficient estimates of all nodewise regressions
• time is the time spent on each regression in second

Help and Support

For questions, issues or feature requests please reach out to Haodong: [e-mail].

Citation

If you use the software, please cite our paper: Liu, J., Wang, H. Sun, W., & Liu, Y. (2021). Prioritizing Autism Risk Genes using Personalized Graphical Models Estimated from Single Cell RNA-seq Data

Reference

Liu, J., Wang, H. Sun, W., & Liu, Y. (2021). Prioritizing Autism Risk Genes using Personalized Graphical Models Estimated from Single Cell RNA-seq Data

hwang655/HUG documentation built on Jan. 31, 2021, 2:54 a.m.