Fig. 4
From: Theoretical characterisation of strand cross-correlation in ChIP-seq
a Comparison between the estimated w with MACS2 and the FWHM of the PyMaSC MSCC distributions (ENCODE A549 dataset). The solid line shows the result of simple linear regression without the intercept term for the TFs. b A log-log plot between the estimated S/N (\(\hat {\alpha }\)) and the numbers of called peaks (\(\hat {n}\)) in the ENCODE A549 dataset. c Spearman’s correlation between each strand cross-correlation-based metric and \(\hat {\alpha }\) (A549 and additional histone-mark dataset). Significant correlations are marked by asterisks (P-value < 0.05). d Comparison of robustness for sequencing depth. Each sample was down-sampled to 5 M, 10 M, 25 M, 50 M and 100 M (if available) mapped reads. The y-axis is normalised into the relative ratio: for each metric and each sample, scores are normalised against the value at all available mapped reads. Dashed lines indicate the ENCODE criteria of usable reads (20 million for TFs and narrow histone marks and 45 million for broad histone marks)