Identifying Non-random Patterns from Gene Expression Profiles
There has been considerable interest in identifying biologically relevant genes from temporal microarray gene expression profiles using linear and nonlinear measures. The present study uses two distinct approaches namely: classical order zero-crossing count (ZCC) and Lempel-Ziv (LZ) complexity in identifying non-random patterns from temporal gene expression profiles. While the former captures the linear statistical properties of the time series such a power-spectrum, the latter has been used to capture nonlinear dynamical properties of gene expression profiles. The results presented elucidate that ZCC can perform better than LZ in identifying biologically relevant genes. The robustness of the findings are established on the given gene expression profiles as well as their noisy versions. The performance of these two techniques is demonstrated on publicly available yeast cell-cycle gene expression data. A possible explanation for the better performance of the ZCC over LZ complexity may be attributed to inherent cyclic patterns characteristic of the yeast cell-cycle experiment. Finally we discuss the biological relevance of new genes identified using ZCC not previously reported.
KeywordsGene expression Time series Zero-crossing count Lempel-Ziv complexity
- 5.Kedem, B.: Time Series Analysis by Higher Order Crossings. IEEE Press, Los Alamitos (1994)Google Scholar
- 9.Proakis, J.G., Manolakis, D.G.: Digital Signal Processing, Principles Algorithms and Applications. Prentice-Hall, Englewood Cliffs (1996)Google Scholar
- 10.Papoulis, A., Pillai, S.U.: Probability, Random Variables and Stochastic Processes, 4th edn. McGraw-Hill, New York (2002)Google Scholar
- 17.Rasband, N.: Chaotic Dynamics of Nonlinear Systems. Wiley-Interscience, Chichester (1997)Google Scholar