Skip to content

Commit c584b4d

Browse files
WPZgithubWPZgithub
committed
Update
1 parent d596e9a commit c584b4d

9 files changed

+106
-55
lines changed

‎ConMod.m

+5-7
Original file line numberDiff line numberDiff line change
@@ -3,23 +3,21 @@
33
% modules in multiple networks
44
%
55
% INPUT:
6-
% multiNetworks: a cell contains multiple networks, each is
7-
% presented by a sparse matrix or a full matrix with N nodes
6+
% multiNetworks : a cell contains multiple networks, each of which is
7+
% presented by edgelist format or a full matrix
8+
% with N nodes
89
% N: the number of all nodes
910
% K: the number of hidden factors
10-
% lambda: a vector containing the parameters for balancing the relative
11+
% lambda: a vector which contains the parameters for balancing the relative
1112
% weight among different views
1213
% xita: the parameter for selecting nodes
1314
% maxIter: the maximum number of iterations for multi-view NMF
1415
%
1516
% OUTPUT:
16-
% modulesfinal: a cell contains the final conserved modules
17+
% modulesfinal: a cell which contains the final conserved modules
1718
%
1819
% Peizhuo Wang ([email protected])
1920

20-
%% parameters
21-
22-
2321
%% Calculting the feature matrices
2422
disp('Calculating the strengh matrix and the uniformity matrix...')
2523
[Strength, Distribution] = featureNets(multiNetworks, N);

‎README.md

+83
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,83 @@
1+
# ConMod
2+
3+
This algorithm is used for identifying conserved functional modules in multiple networks, as described in:
4+
> Feature related multi-view nonnegative matrix factorization for identifying conserved functional modules in multiple biological networks. Peizhuo Wang, Lin Gao, Yuxuan Hu and Feng Li. BMC Bioinformatics, 2018, 19(1): 394.
5+
6+
This algorithm is implemented primarily in Matlab 2015b.
7+
8+
## Usage
9+
10+
### Input Format
11+
12+
The code takes a series of networks as an input. These networks must be stored in a variable of type *cell* in matlab. Each network can be represented by the following two types:
13+
14+
1. adjacency matrix
15+
2. edge list, e.g:
16+
```
17+
1 2 0.62
18+
1 3 0.88
19+
2 9 0.14
20+
...
21+
```
22+
23+
These codes are for generating synthetic datasets:
24+
25+
- `syn_dataset_common.m`: Conserved modules have the same size and are common to a given set of networks.
26+
- `syn_dataset_overlap.m`: Conserved modules are present only in a subset of networks and they are the overlapping parts of specific modules across different networks.
27+
28+
### Main functions
29+
30+
- `ConMod.m`: The implementation of the ConMod algorithm.
31+
```
32+
function modulesfinal = ConMod(multiNetworks, N, K, lambda, xita, maxIter)
33+
% INPUT:
34+
% multiNetworks: a cell contains multiple networks, each of which is presented by edgelist format or a full matrix with N nodes
35+
% N: the number of all nodes
36+
% K: the number of hidden factors
37+
% lambda: a vector which contains the parameters for balancing the relative weight among different views
38+
% xita: the parameter for selecting nodes
39+
% maxIter: the maximum number of iterations for multi-view NMF
40+
%
41+
% OUTPUT:
42+
% modulesfinal: a cell which contains the final conserved modules
43+
```
44+
45+
- `featureNets.m`: Compute two feature matrices which characterize the multiple networks.
46+
```
47+
function [Strength, Participation] = featureNets(multiNetworks, N)
48+
% INPUT:
49+
% multiNetworks: a cell contains multiple networks, each is presented by a sparse matrix or a full matrix with N nodes
50+
% N: the number of all nodes
51+
%
52+
% OUTPUT:
53+
% Strength : N x N matrix for Connection Strength
54+
% Participation : N x N matrix for Participation Coefficient
55+
```
56+
57+
- `multiViewNMF.m`: Multi-view non-negative symmetric matrix factorization.
58+
```
59+
function [H, Hc, objValue] = multiViewNMF(X, K, lambda, maxIter)
60+
% INPUT:
61+
% X: a cell which contains symmetric matrices
62+
% K: the number of hidden factors
63+
% lambda: a vector which contains the parameters for balancing the relative weight among different views
64+
% maxiter: the maximum number of iterations
65+
%
66+
% OUTPUT:
67+
% H: a cell containing factor matrices for all views
68+
% Hc: the result consensus factor matrix
69+
% objValue: the value of objective function
70+
```
71+
72+
- `moduleNodesSelection.m`: Assigning the module members by a soft node selection procedure and then truing the modules to obtain more accurate results
73+
```
74+
function modulesFinal = moduleNodesSelection(Hc, xita)
75+
% INPUT:
76+
% Hc: the consensus factor matrix
77+
% xita: the parameter for selecting nodes
78+
%
79+
% OUTPUT:
80+
% modulesFinal: a cell which contains the final result modules
81+
```
82+
83+
If you have any questions, please contact `[email protected]`.

‎README.txt

-30
This file was deleted.

‎SNMFforView.m

+1-1
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,7 @@
55
% X: the adjacency matrix of a network
66
% Hc: initialization for consensus factor matrix
77
% H: initialization for factor matrix of each view
8-
% lambda: a vector containing the parameters for balancing the relative
8+
% lambda: a vector which contains the parameters for balancing the relative
99
% weight among different views
1010
% MaxIter: the maximal number of iterations for alternating minimization
1111
% epsilon: the convergence parameter

‎evaluation.m

-1
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,6 @@
1010
% FPR: the False Positive Rate
1111
% Accuracy:
1212
% MCC: the Matthews Correlation Coefficient
13-
% I: Confusion Matrix
1413
%
1514
% Peizhuo Wang ([email protected])
1615

‎featureNets.m

+5-5
Original file line numberDiff line numberDiff line change
@@ -2,8 +2,8 @@
22
% Compute two feature metrices from multiple networks
33
%
44
% INPUT:
5-
% multiNetworks : a cell contains multiple networks, each is
6-
% presented by a sparse matrix or a full matrix
5+
% multiNetworks : a cell contains multiple networks, each of which is
6+
% presented by edgelist format or a full matrix
77
% with N nodes
88
% N : the number of all nodes
99
%
@@ -19,7 +19,7 @@
1919
Strength = zeros(N);
2020
temp = zeros(N);
2121
A = zeros(N);
22-
if (m <= 3) % Sparse matrix format
22+
if (m <= 3) % Edgelist format
2323
for k = 1:network_count
2424
theMatrix = multiNetworks{k};
2525
[edge_count, col_count] = size(theMatrix);
@@ -73,9 +73,9 @@
7373
Participation = (network_count/(network_count-1)) * (1-(temp./(A.^2)));
7474
Participation(isinf(Participation)) = 0;
7575
Participation(isnan(Participation)) = 0;
76-
Participation = Participation - diag(diag(Participation));
76+
Participation = Participation - diag(diag(Participation)); % The diagonal is 0
7777

7878
Strength = A./network_count;
79-
Strength = Strength - diag(diag(Strength));
79+
Strength = Strength - diag(diag(Strength)); % The diagonal is 0
8080

8181
end

‎main_run.m

+1-1
Original file line numberDiff line numberDiff line change
@@ -44,7 +44,7 @@
4444
%
4545
% % Selecting nodes from the consensus factors
4646
% xita = 1.5;
47-
% modules_final = modulesTruing( Hc, xita );
47+
% modules_final = moduleNodesSelection( Hc, xita );
4848
% runtime = toc;
4949
% disp(['Running time: ', num2str(runtime), ' sec.'])
5050

‎moduleNodesSelection.m

+5-4
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,13 @@
11
function [ modulesFinal ] = moduleNodesSelection( Hc, xita )
2-
% A soft node selection procedure from the consensus factors to assign the module members
3-
% and then truing the modules to obtain more accurate results
2+
% Assigning the module members by a soft node selection procedure
3+
% and then truing the modules to obtain more accurate results
44
%
55
% INPUT:
66
% Hc: the consensus factor matrix
77
% xita: the parameter for selecting nodes
88
%
99
% OUTPUT:
10-
% modulesFinal: a cell contains the final result modules
10+
% modulesFinal: a cell which contains the final result modules
1111
%
1212
% Peizhuo Wang ([email protected])
1313

@@ -26,7 +26,7 @@
2626

2727
for i = 1:size(HPI, 1)-1
2828
for j = (i+1):size(HPI, 2)
29-
if HPI(i,j)>0.5
29+
if HPI(i,j)>0.5 % merge these two modules
3030
[Y, I] = max([moduleSignal(i), moduleSignal(j)]);
3131
if I == 1
3232
modulesFinal{j} = [];
@@ -43,6 +43,7 @@
4343
end
4444
end
4545

46+
% Only modules with no less than 5 nodes are kept
4647
i = 1;
4748
while i ~= length(modulesFinal)+1
4849
if isempty(modulesFinal{i}) || (length(modulesFinal{i})<5)

‎multiViewNMF.m

+6-6
Original file line numberDiff line numberDiff line change
@@ -2,14 +2,14 @@
22
% Multi-View Non-negative symmetric Matrix Factorization
33
%
44
% INPUT:
5-
% X: a cell containing symmetric matrices
5+
% X: a cell which contains symmetric matrices
66
% K: the number of hidden factors
7-
% lambda: a vector containing the parameters for balancing the relative
7+
% lambda: a vector which contains the parameters for balancing the relative
88
% weight among different views
99
% maxiter: the maximum number of iterations
1010
%
1111
% OUTPUT:
12-
% H: a cell containing factor matrices for all views
12+
% H: a cell which contains factor matrices for all views
1313
% Hc: the result consensus factor matrix
1414
% objValue: the value of objective function
1515
%
@@ -86,9 +86,9 @@
8686
obj_consensus = norm(H{i} - Hc, 'fro')^2;
8787
obj = obj + obj_body + lambda(i)*obj_consensus;
8888

89-
errX = mean(mean(abs(obj_body)))/mean(mean(X{i}));
90-
errH = mean(mean(abs(obj_consensus)))/mean(mean(H{i}));
91-
err = errX + errH;
89+
% errX = mean(mean(abs(obj_body)))/mean(mean(X{i}));
90+
% errH = mean(mean(abs(obj_consensus)))/mean(mean(H{i}));
91+
% err = errX + errH;
9292
end
9393

9494

0 commit comments

Comments
 (0)