Control issues
Extraction of features from a whole chromosome is slower than ideal, I am attempting to accelerate the process by maintaining the bigwig files open as lists (very memory intensive).
From control dataset (\ref{631482} , Mouse dataset CH12.LX) we see that Mg tags are reasonably good at determining the early and late RT regions, which is weird as Liu had the same traits + H3K4me1 and POLR2A, so we would expect Liu tags model to be better than Ng's.