Fig 5. Statistical test for binding region detection.

We executed a right-tailed t-test for the number of covering binding sites from the convolution results with a null distribution, which was constructed from the randomly generated convolution results in the sc-PDB database consisting of 7,179. Because each sc-PDB test has many windows, we selected the most significant p-values adjusted by the Benjamini-Hochberg procedure and examined whether they were significant at levels of 1%, 5% and 10%. The results showed that 14.6%, 30.3% and 42.2% of sc-PDB entries were significantly enriched, respectively (A-C).