Steps to apply sucessfuly MC_MCL in MATLAB (tested using cygwin in a windows 10 environment).
- for windows users: Install cygwin, a collection of GNU and open source tools which provide functionality like in linux.
- Install C/C++ compilers in the cygwin/linux environment (required for step 3).
- Install MCL following the instructions as in https://micans.org/mcl/
- MC_MCL should be ready to use. Its I/O are:
Inputs:
x: matrix data where in the rows are the samples and in the columns the features.
C: number of clusters that the user is trying to find.
dist: number that represent the distance applied on "x" to construct the graph [1] apply the MC matrix created with euclidean distance [2] apply the MC matrix created with correlation distance [3] apply the MC matrix created with correlation distance default: 1
cygwin_path: cygwin path to run MCL (folder where cygwin was installed)
factor: the option for the MC distance: [1] nothing [2] sqrt [3] log(1+x) default: 1
cutoff: threshold in the similarity of the samples used to compute the graph. The higher the cutoff it is, the more number of components the graph may have. default: the program calculate for itselfs the higher cutoff where the graph contains 1 component.
max_nc: maximum number of components of the graph for choosing the cutoff default : 1
mod: perform modality: [1] choosing cutoff automatically, matrix-dataset goes from 0 until 1-cutoff. [2] sparcifying network using cutoff selected by the user default: 1
Outputs:
comm: vector of size of the lengths of the rows of x containing the associated clusters to the samples.
nc: number of components of the network introuced to MCL.