Low rank approximation based multi-omics data clustering (LRAcluster) is a new method to discover molecular subtypes by detecting the low-dimensional intrinsic space of high-dimensional cancer multi-omics data. The low-rank constraint is the core to generate the low-dimensional representation of the original data. And the convexity of the regularized likelihood function provides efficient gradient-descent algorithm for optimization. Extensive experiments show that LRAcluster is computationally efficient with high accuracy and thus suitable for large-scale cancer multi-omics studies.
Source Codes: LRAcluster_1.0.tgz
Windows Binary: LRAcluster_1.0.zip
Gu's Group Homepage