Impact of the choice of normalization method on molecular cancer class discovery
using nonnegative matrix factorization

Haixuan Yang, Cathal Seoighe

Updated on 01 June 2016

The MATLAB code for the algorithm described in the paper can be downloaded here

Acknowledge:

1. We used the function AccMeasure downloaded from http://www.cad.zju.edu.cn/home/dengcai/Data/Clustering.html.

2. We used the code and datasets downloaded from
http://www.broadinstitute.org/cgi-bin/cancer/publications/pub_paper.cgi?paper_id=89

Main functions:

· pNMF.m: The implementation of Algorithm 1.

Usage: type

· help pNMF

or see example.m

In the following, pNMF is not called directly because, for the purpose of easy pair-wise comparisons and of less running time, the same NMF solutions by the Basic NMF should be repeatedly used by all normalization methods.

To repeat some results in Table 1 run

· exp_BasicSetting_VariousNorm.m

To repeat results in Fig. 1, set TYPE='KL' and run

· exp_ALL_AML_Part.m: for Leukemia.

· exp_CPG1_part.m: for CNS.

· exp_Medulloblastoma_part.m: for Medulloblastoma.

· Report_Part.m

To repeat results in Fig. 2, run

· exp_add_noise.m

· Report_exp_add_noise.m

To repeat results in Fig. 3, run

· exp_ALL_AML_Conconsensus_nobootstrap.m

· exp_CPG1_Conconsensus_nobootstrap.m

· exp_Medulloblastoma_Conconsensus_nobootstrap.m