Independent arrays or independent time courses for gene expression time series

We apply three different independent component analysis (ICA) methods, spatial ICA (sICA), temporal ICA (tICA), and spatiotemporal ICA (stICA), to gene expression time series data and compare their performance in clustering genes and in finding biologically meaningful modes. Only spatial ICA was applied to gene expression data previously (Lee, S. and Batzoglou, S., Advances in Neural Information Processing Systems, vol.16, 2004; Liebermeister, W., Bioinformatics, vol.18, no.1, p.51-60, 2002). However, in the case of yeast cell cycle-related gene expression time series data, our comparative study reveals that tICA outperforms sICA and stICA in the task of gene clustering and stICA finds linear modes that best match the cell cycle.