Predictive Genome Model

Predictive Genome Model is a predictive and transferable model to simulate the structure and dynamics of chromosomes.

Dependencies

The package for Predictive Genome Model is mainly written in Python. The LAMMPS package is written in C++.

Visualization of chromosome structures

The information of trajectories stored in .dcd format can be visualized using VMD http://www.ks.uiuc.edu/Research/vmd/ or any other software that is capable of loading .dcd files and available online. Refer to ./PredictiveGenome/contactMap/README.md for more details on visualization of the simulated chromosome structure.

Installation

Download the Predictive Genome Model source package by running the following command:

git clone https://github.com/qiyf/PredictiveGenome.git

or download the zip file with the link:

https://github.com/qiyf/PredictiveGenome/archive/master.zip

The LAMMPS package can be downloaded by running the following command:

git clone -b stable https://github.com/lammps/lammps.git

or with other links from the official site:

http://lammps.sandia.gov/download.html

After download both of the packages, unpack both Predictive Genome Model and LAMMPS packages under the same directory. Then the modified potential files can be connected to the LAMMPS source code by executing the following command (rename the LAMMPS folder as ./lammps/ or change the name of the downloaded LAMMPS folder in the file ./PredictiveGenome/lammpsCode/src/connectLAMMPSPotential.sh):

./PredictiveGenome/lammpsCode/src/connectLAMMPSPotential.sh

The corresponding potential files should be generated under ./lammps/src/.

To compile the parallel version of LAMMPS, go to the folder ./lammps/src/, and type the command:

make clean-all
make openmpi

To include the packages, type the command (e.g. class2 package):

make yes-class2

For more details, please refer to the instrucion on the official site:

http://lammps.sandia.gov/doc/Section_start.html.

Getting start to build the first simulation

Enter the ./PredictiveGenome/ folder to execute the following steps.

The 1D input of the model consists of two parts: chromatin states and CTCF-binding sites, which will be illrustrated in the following:

Chromatin states derived from the underlying combinatorial patterns of key histone marks.

To construct the chromatin states for GM12878, chromosome 1, run the following command:

./example/genChromState.sh

The generated chromatin states file is located at ./chromatinStates/model_input/Gm12878/Gm12878_chr1_chromatin_states.txt

To process other chromosomes and cell types, go to folder ./chromatinStates/ and refer to ./chromatinStates/README.md for details.

CTCF-binding sites derived from ChIP-Seq experiments.

To construct the CTCF-binding sites for GM12878, chromosome 1, run the following command:

./example/genCTCFbinding.sh

The generated chromatin states file is located at ./ctcfBindingSites/processedCTCF/model_input/

To process other chromosomes and cell types, go to folder ./ctcfBindingSites/ and refer to ./ctcfBindingSites/README.md for details.

Run the simulation

To initialize the simulation for GM12878, chromosome 1, run the following command:

./example/runSimulation.sh [-p ncpu] [-i partition] [-t simulation_time] [-l Lammps_dir]

There would be an option for the availability of the computing cluster. If cluster resource is available, specify the number of cpu, the name of the partition available on the cluster, and the time limit of the node.

To process other chromosomes and cell types, go to folder ./runSimulation/ and refer to ./runSimulation/README.md for details.

Contact map

To calculate the contact map for GM12878, chromosomes 1 after running the simulation, run the following command:

./example/calContactMap.sh [-n runnumber] [-u username] [-i partition]

There would be an option for the availability of the computing cluster. If cluster resource is available, specify the number of parallel running be included to calculate the contact map, the username on the cluster, and the name of the partition available on the cluster.

To process other chromosomes and cell types, go to folder ./contactMap/ and refer to ./contactMap/README.md for details.

The experimental Hi-C data is available with GEO accession number GSE63525 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE63525.

The visualization of the simulated chromosome structure ensemble is also available. Please refer to the Visualization of chromosome structures section on where to download the VMD software, and ./contactMap/README.md for details about the steps on visualizing the structures.