MKerW-A is a novel multi-view spectral clustering framework to integrate different omics data types measured from the same subjects by treating each omic data type as a different informative representation between patients. MKerW-A learns the weight of each data type as well as a similarity measure between patients via a non-convex optimization framework. It solves the proposed non-convex problem iteratively using the ADMM algorithm.

• generate_sim_matrices2.m : Contruct multiple similarity matrices using Gaussian kernels (Step 1).

• clus_sim_update2_2HW.m : Main MKerW-A algorithm solving optimization problem (Step 2).

• clus_sim_update0_2HW.m : Algorithm solving embedded ADMM algorithm in Step 2.

• Example (BRCA) : Simple example file of MKerW-A using BRCA cancer.

Please follow the links to reproduce the clustering results of TCGA data sets

• Clustering analysis for each cancer type : Generate clustering results for each of the 22 cancer types.

• Clustering analysis for identifying 22 cancer types : Generate clustering results related to identifying 22 cancer types.

• Analysis on target cluster number : Choose target cluster number.

• Sensitivity analysis : Sensitivity analysis with respect to changes of reglarization parameters.

• Sensitivity test w.r.t. additive noise : Sensitivity analysis with respect to additive noise.

Note Most of the simulations and TCGA data applications were implemented on an Apple MacBook Pro (2.7 GHz, 8 GB of memory) using the MATLAB 2016b.

• All the functions used in the proposed algorithm MKerW-A are located in the directory "Main_Code".

• All the other supplementary files are located in the directory "Other_codes".

• All the codes generating figures presented in the manuscript are located in the directory "Generating_Figures".

• All the resulting files (e.g. .MAT and .eps) are located in the directory "Results_files".

The 22 TCGA cancer data sets saved in the matlab file can be obtained from the dropbox directory all_data.mat.

Specifically, all_clin includes clinical information of the patients from the 22 cancer types.

• all_exp is the RNA data for all the patinets.

• all_mirna is the MicroRNA data for all the patinets.

• all_cna is the CNA data for all the patinets.

• all_pat is the index vector of patients indicating corresponding cancer types (See the first column of all_clin for the original cancer name).

We provide MATLAB implementations of MKerW-A in the MKerW-A branch. The 22 TCGA cancer data sets saved in the matlab file can be downloaded from the dropbox directory all_data.mat.

• Seyoung Park and Hongyu Zhao

  Department of Biostatistics, School of Public Health, Yale University

• [email protected]

This project is licensed under the MIT License.