1. Judgment algorithm for periodicity of time series data based on Bayesian information criterion.
- Author
-
Tominaga D and Horimoto K
- Subjects
- Animals, Bayes Theorem, Mice, Algorithms, Circadian Rhythm physiology, Decision Support Techniques, Gene Expression Profiling methods, Multigene Family physiology, Proteome metabolism
- Abstract
Judgment periodicity of biological time series data is important and done widely to find the circadian expression of genes, monthly change of hormones, etc. To keep complete reproducibility of judgment is a problem because popular judgment methods such as curve fitting, Fourier analysis, etc. need judgment criteria determined by analysts considering experimental conditions and results (level, S/N, distribution, etc.) based on their experience. Judgment results are often affected by analysts' subjects. Reproducible criterion determination is therefore strongly needed. We propose introducing the information criterion to replace analysts' criteria. A judgment algorithm by combining Bayesian information criterion (BIC) and discrete Fourier transform (DFT) has been developed and has proved its ability through application to mice microarray data and finding of circadian genes. Our method, named "Piccolo", shows higher sensitivity than the simple DFT (without BIC) method with reproducibility, and can be fully automated.
- Published
- 2008
- Full Text
- View/download PDF