Nonnegative tensor factorization for clustering genes with time series microarrays from different conditions: A case study

Gene clustering analysis with microarray data plays an important role in understanding gene function and biological process. This paper considers clustering genes with time series microarrays from multiple and different studies in which data lies in tensor space of genes × time-points × studies. Traditional clustering methods deal the data in matrix space of genes × time-points where time-points are combined from different studies. We compare nonnegative matrix factorization (NMF) and nonnegative tensor factorization (NTF) for clustering genes. The experimental results show that NTF outperforms NMF in determining the number of clusters and achieving high clustering accuracy.