Biclustering-SSVD

Implementation of the sparse singular value decomposition algorithm (Lee et al., 2010) to be used for Biclustering; Duke STA663 Final Project

The Sparse Singular Value Decomposition for Biclustering notebook contains the final, optimized version of the algorithm from Lee et al., 2010, whose code is publically available in matlab and R (see ssvd.R, ssvd.m, and thresh.m). This SSVD algorithm is meant to be used for biclustering applications and analysis. The Sparse SVD First Pass notebook is a direct portiing of the code from matlab to python, but has not been optimized, and such runs much slower. The First Pass algorithm runs at approximately 18 minutes, whereas the optimized version run at approximately 40 seconds. Data.txt is the application example dataset (the genes example) provided by the authors of the paper. The paper, Lee et al., 2010, can be found at ssvd.pdf, and the supplemental to the paper can be found at LeeShenHuangMarron09-sup.pdf.

Installation Instructions

1. Either git clone this repository in order to obtain all the files, or
2. Execute the following in terminal pip install Biclustering-SSVD in order to download ssvd_fast.py and ssvd_original.py.