INGOR is an implementation of the algorithms for estimating Bayesian networks with B-spline nonparametric regression model, presented mainly in Imoto et al. (2002) and Tamada et al. (2011).

Visit https://ytlab.jp/clinfo/ingor/ for the latest information and documents for INTERNAL USE.

NOTE: THIS IS THE OKUNO LAB INTERNAL VERSION. DO NOT DISCLOSE ANYTHING WITHOUT PERMISSION. BE CAREFULL FOR HANDLING THE SOURCE CODE. OPEN SOURCE VERSION IS DEVELOPED IN 'oss' BRANCH.

NOTE: ECv CALCULATION IS PATENT PENDING. KYOTO TLO HAS THE COMMERCIAL RIGHTS OF THIS SOFTWARE.

INGOR works with general linux systems, macOS, Windows and R-CCS supercomputer fugaku.

INGOR WILL be distributed under The 3-Clause BSD License except for ECv calculation code.

Pre-compiled executable binaries

Under preparation.

How to build

1. Clone (or pull/checkout) the `ingor` repository on Github.

1.0 Preparation

You require a C compiler such as GCC and a BLAS/LAPACK library to compile INGOR from the source code. You can use Intel C compiler and its math library MKL as a BLAS/LAPACK library. MPI implementation is also required for the NNSR algorithm. This is confirmed to work with OpenMPI, Intel MPI Libarary, and Fugaku MPI libary.

1.1 Clone the specific release

If you want the specific release of INGOR, use git clone -b RELASE_TAG_NAME command where RELEASE_TAG_NAME represents the release tag name such as r1.0.0.

Via HTTPS

$ git clone -b RELEASE_TAG_NAME https://github.com/ytlabjp/ingor.git

Via SSH

$ git clone -b RELEASE_TAG_NAME git@github.com:ytlabjp/ingor.git

1.2 Clone the latest update

If you want to clone the current repository but not yet released with the specifiec release number, simply use the git clone command without the -b option.

Via HTTPS

$ git clone https://github.com/ytlabjp/ingor.git

Via SSH

$ git clone git@github.com:ytlabjp/ingor.git

2. Prepare your own settings in `Makefile.local`

Currently, INGOR does not provide the configure script, which inspects your environment requred for the compilation of the source code. Instead, prepare your Makefile.local file depending on your environment for your self.

cd ingor vi Makefile.local

There are some prepared Makefile.local files for general Linux, Linux with the Intel compiler, MacOS and Fugaku. If you use them, copy one of Makefile.* as Makefile.local where * represents the system you are using. See these files for details.

3. Build

Just type make in the ingor directory.

make

The binary executable file ingor will be generated.

To enable MPI-based algorithms such as the NNSR algorithm, use

make USE_MPI=1

command. The MPI algorithm-enabled binary will be generated. Be careful that, for some environments such as Fugaku, make command with its prepared Makfile generates the MPI-enabled binary by default.

4. Install and Execute

The generated binary can be executed directly like other commands.

./ingor.0.X.0

If you are a superuser of your system, you can install INGOR under, for example, /usr/local/bin (by default).

sudo make install

For some environments such as Fugaku, the prepared Makefile generates the binary for computation nodes, not for log-in environments. These are different and the single binary does not run on both environments.

The MPI-enabled binary needs to be executed via mpiexec or mpirun command.

See the online line documentation at https://ytlab.jp/clinfo/ingor/ for how to use the software.