MCL_lite

MCL_lite is a memory-efficient implementation of Markov Clustering Algorithm for large-scale networks. In our tests, MCL_lite can cluster a network with 2 billion edges on a desktop with 16GB RAM while the original MCL needs more than 100GB RAM.

Requirement

Currently, MCL_lite only support 64-bit linux operation system. Make sure that you have the following installed:

1. Python2.7 (Recommend Anaconda)

   1. numpy(version>=1.15.4)

   2. scipy

   3. sklearn

   4. numba

   5. Install dependency packages via pip:

      $ pip install scipy numpy scikit-learn

   6. Install dependency packages via conda:

      $ conda install scipy numpy scikit-learn

Download

$ git clone https://github.com/Rinoahu/MCL_lite

Usage

$python MCL_lite/mcl_sparse.py -i foo.xyz -I 1.5 -o foo.mcl -a 8 -m 8 -d t

-i: input network in 3 column tab-delimited format. The fist and second columns stand for nodes and third column stands for weight between these two nodes. For example:

N0	N1	1.5
N0	N2	1.2
N0	N3	1.1
...

-I: float. inflation parameter for mcl

-a: int. cpu number

-o: str. name of output file

-d: t|f. is the graph directed? Default is t(true)

-g: int. how many gpus to use for speedup. Default is 0

-r: t|f. resume the work after crash. Default is f(false)

-m: int. memory usage limitation. Default is 4GB