netalign : Network Alignment codes

Mohsen Bayati, David F. Gleich, Amin Saberi, Ying Wang

Supporting codes for the manuscript: Message Passing Algorithms for Sparse Network Alignment.

This package contains datasets and programs to solve network alignment problems written in the Matlab language.

These codes are research prototypes and may not work for you. No promises.


Version 1.1 : last updated 2013-01-22, but this should be the same as the old 20110227.tar.gz version. * See an older version here netalign-online-1.0.tar.gz

{: .nobullets}


The package is organized by directory

datasets for the experiments
script files that use and evaluate the algorithms
code for all the main algorithms
a historically named directory for our data files


With a recent version of Matlab (2008b used for development), the best way to use the package is to open Matlab, navigate to the matlab directory, and then execute the following commands to solve the network alignment problem for Figure 1.

% load A, B, and L for figure 1
>> load('../data/example_overlap.mat');

% create S, w, and a variant of L from graph A, B, and L
>> [S,w,li,lj] = netalign_setup(A,B,L);

% use S, w, L, and alpha=0, beta=1 and the bp algorithm
>> x = netalign_bp(S,w,0,1,li,lj);

% x is just a heuristic, so we need to round it
>> [ma mb mi overlap weight] = mwmround(x,S,w,li,lj);
% [ma mb] give pairs of matched vertices 
% mi is the binary matching indicator for L (as li, lj)
% overlap and weight are the overlap and weight objectives

If you want to get more information about how to use the tools, see the matlab/demo.m script.


All of the algorithms and codes are contained in the matlab directory.

|Algorithm |Code |Source | |:--------------------|:---------------------|:----------------| |IsoRank |full_isorank.m |Singh et al. 2007| |SpaIsoRank |isorank.m | | |NetAlginBP |netalignbp.m | | |Exact enumeration |netalign_exact.m | | |NetAlignMR |netalignmr.m |Klau 2009| |LP |netalign_lp_prob.m |Klau 2009| |Lagrangean LP |netalign_llp.m |Klau 2009|


|Dataset |File |Source | |:------------------|:------------------------------------|:------------------| |lcsh2wiki-small |private_data/lcsh2wiki-small.mat | | |lcsh2wiki-full |private_data/lcsh2wiki_full.mat | | |dmela-scere | | Singh et al. 2008 | |musm-homo | | Klau 2009 |


These data sets only come with S, w, li, and lj. The original datasets are considerably larger (hundreds of megabytes). Please contact us if you want them. None of the experiments require them.


These datasets are distributed with IsoRank in the file .

Use the python program experiments/bioinfo/ to generate dmela-scere.smat dmela.smat and scere.smat which are L, A, and B, respectively. The program readSMAT.m will load these files into Matlab.

See for more about IsoRank.

After converting the data to smat with the above script, execute

A = readSMAT('dmela.smat');
B = readSMAT('scere.smat');
L = readSMAT('dmela-scere.smat');
[S,w,li,lj] = netalign_setup(A,B,L);
save '../../data/dmela-scere.mat' A B L S w li lj

to save the data for use with the experiments.


These datasets are distributed with Natalie in the file .

Use the perl script experiments/bioinfo/ in the to extract the files and then load_natalie.m to read the data in Matlab.

In the experiments/bioinfo directory, execute

[S,w,li,lj] = netalign_setup(A,B,L);
save '../../data/natalie_graphs.mat' A B L S w li lj

to save the data for use with the experiments.


|Experiment|Description|Figure| |:------------------|:------------------------------------|:------------------| |misc/figure_2.m | Generate figure 2 | Figure 2 | |bioinfo/bioinfo.m | Playing with musm-homo and dmela-scere datasets | | |evaluation/evaluate_all_problems.m | Generate results for figures 7 and 8 and table 4, plot with evaluation_plots.m and evaluation_table.m | Figures 7, 8, Table 4 | |grid_test/grid_test.m | Synthetic grid experiments, plot with plot_grid_test.m | Figure 6 | |lcsh2rameau/rameau2lcsh.m| Rameau and LCSH experiments, make table with rameau2lcsh_table.m | Table 4 | |powerlaw/powerlaw.m | Synthetic power law experiments, plot with plot_powerlaw.m | Figure 6 | |rounding/round_test.m | Variations on rounding schemes | |