RUN_Cross_Validation.m

%--------------------------------------------------------------------------

%
% Author: 
%Junjun Zhang and Minzhu Xie
% 
% Purpose:
% Perform drug-target interaction prediction in a cross-validation setting
% to estimate prediction performance of various prediction methods.
% (DEFAULT: 5 repetitions of 10-fold cross validation)
%

clear   % clear workspace
clc     % clear console screen

diary off;  diary on;   % to save console output

%--------------------------------------------------------------------------

%*************************%
%* Adjustable Parameters *%
%*************************%

% The location of the folder that contains the data
path='data\';

% the different datasets
datasets={'e','ic','gpcr','nr'};

% CLASSIFIER -------------------------------
classifier='iPALMDLMF';
% ------------------------------------------

% Parameters and Options -------------------
%WKNKN
use_WKNKN = 0;      % 1=yes, 0=no
K = 5;              % number of K nearest known neighbors
eta = 0.7;          % decay rate (also used by WNN in RLS-WNN)


%weight matrix W (used by WGRMF and CMF)
if strcmp(classifier,'wgrmf') || strcmp(classifier,'cmf')
    use_W_matrix = 1;
else
    use_W_matrix = 0;
end
% ------------------------------------------

% CROSS VALIDATION SETTING -----------------
 cv_setting = 'cv_d';    % DRUG PREDICTION CASE
% cv_setting = 'cv_t';    % TARGET PREDICTION CASE

% ------------------------------------------

% CROSS VALIDATION PARAMETERS --------------
m = 5;  % number of n-fold experiments (repetitions)
% m = 1; 
n = 10; % the 'n' in "n-fold experiment"
% n = 5;
% ------------------------------------------

%warning off     % to be used when many unnecessary warnings are being produced

%--------------------------------------------------------------------------

% Terminology:
% Y = Interaction matrix
% Sd = Drug similarity matrix
% St = Target similarity matrix


disp('==============================================================');
fprintf('\nClassifier Used: %s',classifier);
switch cv_setting
    case 'cv_d', fprintf('\nCV Setting Used: CV_d - New Drug\n');
    case 'cv_t', fprintf('\nCV Setting Used: CV_t - New Target\n');
end
if use_WKNKN
    fprintf('\nusing WKNKN: K=%i, eta=%g\n',K,eta);
end
fprintf('\n');

% ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

for ds=[4 3 2 1]
    disp('--------------------------------------------------------------');

    fprintf('\nData Set: %s\n', datasets{ds});

    % LOAD DATA
    [Y,Sd,St,Did,Tid]=getdata(path,datasets{ds});

    % PREDICT (+ print evaluation metrics)
    crossValidation(Y',Sd,St,classifier,cv_setting,m,n,use_WKNKN,K,eta,use_W_matrix);

    disp('--------------------------------------------------------------');
    diary off;  diary on;
end

% ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

disp('==============================================================');
diary off;