3run_PhiMRF

Running PhiMRF

All of the following steps are in pipeline_PhiMRF.sh.

Please read through each step and customize, before running

bash pipeline_PhiMRF.sh

The vignettes in the PhiMRF package has very detailed examples of how to use the package.

Two datasets are required,

• Adjacency matrix of network structure:

  • the processed HiC gene network

  • Example: $homedir/intra/data/chrm1_0.9_mean_TRUE_FALSE_neighbors_trans.rds

• Observed count y:

  • processed RNA-seq count,

  • Example: $homedir/intra/data/y/chrm1_y.rds

Designate home dirctory

The same directory as the set up step.

Example:

homedir=/home/nzhou/hic/rao2014/GM12878_10kb/

Designate quantile and method

quantile=0.9
method="mean"

Create results folders

mkdir $homedir/intra/results/
mkdir $homedir/inter/results

Before you run PhiMRF in a loop...

1. Please read the vignette in the PhiMRF package

2. Please edit the R scripts run_PhiMRF_*.R according to the documentation in PhiMRF.

Arguments to edit include:

• number of total iteration of MCMC

• number of burn-in iterations

• variance of random walk chains

• Parameters for uniform distributions

• etc (see documentation for the function PhiMRF::pmrf())

The R scripts to edit:

• run_PhiMRF_intra.R

• run_PhiMRF_linear.R

• run_PhiMRF_inter.R

3. Each run of PhiMRF for each chromosome (pair) could take minutes to hours, depending on the size of the chromosome (pair), the number of iterations, and your computing power.

Please tune the model first with a single chromosome (chromosome 1 is the largest), before committing to a big loop as shown below. The following loop goes through all 253 pairs of chromosomes, and is likely going to take days.

Run everything in a loop

for chrmA in "${chrms[@]}"
do
	Rscript run_PhiMRF_intra.R $homedir/intra "$chrmA" "$quantile" "$method"
	Rscript run_PhiMRF_linear.R $homedir/intra "$chrmA" 
	for chrmB in "${chrms[@]}"
	do
		Rscript run_PhiMRF_inter.R $homedir/inter "$chrmA" "$chrmB" "$quantile" "$method"
	done
done