Wiki

Clone wiki

ibas / Home

Welcome to the IBAS wiki!

Understanding the output format

The first two lines in your output describe the ROC curve area under the curve (AUC), along with the significance of observing a AUC of that magnitude at random. The AUC and its p-value tell you how well the proteins from your input set predict each other, or, in other words, how well the proteins from your input set are connected to each other.

# AUC:  0.73517
# AUC P-value:  0.001

The next lines of your output starting with 'Model' specify for each of the input genes, the

  1. Gene ID of the gene by the nomenclature specified by the network.
  2. Gene symbol
  3. The network order of the model that most optimally predicts the input gene based on the other genes (can be either first-, or second-order)
  4. The interaction threshold of the model that most optimally predicts the input gene based on the other genes
  5. The AUC of the model that most optimally predicts the input gene
# Model: Gene ID                  Gene Symbol     Order   Threshold       AUC
# Model: ENSG00000165410          CFL2            first   0.25000         0.60655
# Model: ENSG00000143632          TNNT1           second  1.00000         0.73705
# Model: ENSG00000177455          ACTA1           second  0.20000         0.67336
# Model: ENSG00000198807          MC4R            second  0.20000         0.67636
# Model: ENSG00000105048          TPM3            second  0.20000         0.67294
# Model: ENSG00000153922          FTO             second  0.05000         0.74478

The rest of the lines specify for each gene in the network, the

  1. Gene ID of the gene by the nomenclature specified by the network.
  2. Gene symbol
  3. The posterior probability of that gene/protein being a 'phenotype causing gene/protein'
  4. The empirical p-value
  5. The false-discovery rate.
Gene_Inweb_Id             Gene_Symbol     Posterior_Probability   Empirical_P-value       False_Discovery_Rate
ENSG00000198802           CCND2           0.0023                  0.001                   0.01
ENSG00000139220           NARS            0.0021                  0.001                   0.01
ENSG00000143549           LIX1L           0.0012                  0.002                   0.01

BROAD Institute Howto

Running instructions

At the BROAD System you will find an example run here:

/home/unix/tunepers/tools/ibas/example

Please copy and modify the following files to your local IBAS directory:

/home/unix/tunepers/tools/ibas/example/run_ibas.sh
/home/unix/tunepers/tools/ibas/example/ibas.conf 

The first script shows you how run run IBAS through the BROAD queeing system, while the other is an example IBAS config file (the only argument to IBAS).

In the config file there are three lines you should edit:

num_permutation_sets = 1
num_threads = 1
pcp_list_file_path = /home/unix/tunepers/tools/ibas/example/genes.ens

Please also makes sure that the following parameters are used:

verbosity = prog
num_cv_reference_genes = 2000

'num_permutation_sets' specifies the humber of permutations you would like to run. Each permutation is run within a separate thread and takes roughly 1 minute to compute. If your aim is to test whether the genes in your input set are able to predict each other, you should set this number to 0. If your aim is assess whether their mutual ability to predict each other deviates from random expectations, you should run 100 or 1000 permutations.

'num_threads' specifies the number of threads. Note, if you change that number you also need to change it in the run_ibas.sh wrapper script.

'pcp_list_file_path' specifies the phenotype causing proteins that you input to IBAS. For now please specify these in Ensembl format.

Run IBAS:

You run an IBAS analysis by issuing the following commands:

use LSF
use GCC-4.4
use .boost-1.52.0
/home/unix/tunepers/tools/ibas/example/run_ibas.sh

If you only would like to run a small number of permutations (0-10), you could run IBAS in an interactive shell (remember to set 'num_threads = 1')

use LSF
ish
use GCC-4.4
use .boost-1.52.0
/home/unix/tunepers/tools/ibas/kernel/ibas.bin ibas.conf

Compilation instructions

Go ahead and get the code:

git clone https://tunepers@bitbucket.org/HeyHo/ibas.git

If you get a 'SSL certificate problem' error, you could disable SSL certificate before cloning the directory:

git config --global http.sslVerify false
git clone https://tunepers@bitbucket.org/HeyHo/ibas.git

and compile the program:

use LSF
use GCC-4.4 //The instructions from broad had a typo
use .boost-1.52.0
ish
cd kernel
make -f Makefile-x86-broad clean
make -f Makefile-x86-broad

General Installation Howto of development version

Change to directory where IBAS code lives. If you have not checked it out before, use the 'git clone' command shown above. Remember to use the Makefile made for your system (the one below will work on a Ubuntu Linux system):

git checkout devel
git pull
cd kernel
make -f Makefile-x86-broad clean
make -f Makefile-x86-broad

Please also make sure that 'verbosity' is set to 'prog' and that 'num_cv_reference_genes' is set to 2000 in the config file when you run IBAS.

In the case of facing "HTTPS cloning errors", check your git version using

git --version

if its lower than 1.7, and you are installing IBAS on Broad, use /home/unix/tunepers/local/tools/git-master/git instead

Have fun!

Updated