Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Constrained Fourier estimation of short-term time-series gene expression data reduces noise and improves clustering and gene regulatory network predictions

Fig. 2

Results of k-means clustering of raw (gray) and de-noised (red) synthetic expression data. A, B Six synthetic clusters, from each we generated 1000 signals with random additive noise of variance \(\sigma ^2=0.1\) (A) and \(\sigma ^2=0.9\) (B). Fourier approximation of de-noised data that was clustered (red dashed) and Fourier approximation of raw data that was clustered (gray dashed). C, D Monte Carlo of 1000 k-means simulations (see “Methods” section) on the de-noised and raw signals. The histograms describe the distribution of the SSEs for the raw (grey) and the de-noised (red) data. The mean error SSE of Fourier treated genes (\({\bar{SSE}}=1.9\)) was significantly lower (t-test: \(p<0.01\)) than the mean SSE of the untreated genes (\({\bar{SSE}}=3.4\)).The difference in low noise signals (here shown \(\sigma ^2=0.1\)) was also statistically significant (t-test: \(p<0.01\))

Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us