Skip to main content

Table 2 Performance indices on artificial data of half-lives and expression time-courses with different noise amplitudes (C)

From: StaRTrEK:in silico estimation of RNA half-lives from genome-wide time-course experiments without transcriptional inhibition

  \(\alpha _{opt}\) \(q_{\text {max}}\) \(\rho\) pval \(FDR<0.05\)
\(C=0.1\) 10 1 0.90 \(<10^{-309}\) \(99.6\%\)
\(C=0.2\) 10 2 0.86 \(1.3\cdot 10^{-277}\) \(95.2\%\)
\(C=0.3\) 10 2 0.86 \(2.3\cdot 10^{-269}\) \(91.1\%\)
\(C=0.4\) 10 4 0.68 \(8.9\cdot 10^{-123}\) \(90.8\%\)
\(C=0.5\) 10 10 0.11 \(3.9\cdot 10^{-4}\) \(90.8\%\)
  1. Numbers of gene expression profiles (\(m = 1000\)) and of time samples (\(n=6\)) are kept fixed. Legend—\(\alpha _{opt}\): optimal value of the regularization parameter; \(q_{\text {max}}\): error threshold expressed as percentile of the MSE distribution; \(\rho\): Pearson correlation coefficient; pval: p value; FDR: false discovery rate. For each instance of the noise distribution considered, we found a negligible variability of the quality indices, therefore variance is not reported in the table