EBIC - AI-based parallel biclustering algorithm

EBIC is a next-generation biclustering algorithm based on artificial intelligence (AI). EBIC is probably the first algorithm capable of discovering the most challenging patterns (i.e. row-constant, column-constant, shift, scale, shift-scale and trend-preserving) in complex and noisy data with average accuracy of over 90%. It is also one of the very few parallel biclustering algorithms that use at least one graphics processing unit (GPU) and is ready for big-data challenges.

EBIC is mainly implemented in C++11. CUDA with OpenMP used for parallelization.

The latest version of EBIC works also for Big Data.

EBIC is still under active development

License

EBIC is MIT-licensed. Please see the repository license for the licensing and usage information.

Citation

If you happen to use EBIC for mining your data, please cite us using the following BibTex entry:

@article{doi:10.1093/bioinformatics/bty401,
  author = {Orzechowski, Patryk and Sipper, Moshe and Huang, Xiuzhen and Moore, Jason H},
  title = {EBIC: an evolutionary-based parallel biclustering algorithm for pattern discovery},
  journal = {Bioinformatics},
  volume = {},
  number = {},
  pages = {bty401},
  year = {2018},
  doi = {10.1093/bioinformatics/bty401},
  URL = {http://dx.doi.org/10.1093/bioinformatics/bty401},
  eprint = {/oup/backfile/content_public/journal/bioinformatics/pap/10.1093_bioinformatics_bty401/3/bty401.pdf}
}

Installation

EBIC requires CUDA 8.0, installed C++11 environment and OpenMP. We maintain the EBIC installation instructions.

Usage

EBIC can be used only as a command line tool.

In order to build a program simply run: $ make

Check our 'input.txt' data file to see the required input file format. In order to run an example simply type: $ ./ebic -i input.txt

The basic usage of EBIC is: $ ./ebic [OPTIONS]

To override any of default options extra arguments should be added:

Options:

• -i,--input TEXT input file

• -n,--iterations INT number of iterations [default: 5000]

• -b,--biclusters INT number of biclusters [100]

• -x,--overlap FLOAT overlap threshold [0.75]

• -g,--gpus INT number of requested GPUs [1]

• -a,--approx FLOAT approximate trends allowance [0.85]

• -t,--negative-trends INT negative trends [1]

• -l,--log is logging enabled [false]

• -m,--missing_calue INT a parameter substituting a missing value (in order to change the focus of the method)

Examples

Check available options:

$ ./ebic -h

Run EBIC for 10 iterations and return 5 biclusters only:

$ ./ebic -i input.txt -n 10 -b 5

Do not allow negative trends:

$ ./ebic -i input.txt -t 0

Do not allow approximate trends:

$ ./ebic -i input.txt -a 1