Fig. 5

Example of visualisation of atyPeak results in the UCSC genome browser. The results presented here are for the HeLa cell line for ReMap 2018. A darker peak indicates a higher atyPeak score. The annotated BED data files with the corresponding atyPeak scores are available at <https://github.com/qferre/atypeak-files> or as a UCSC browser session at <http://genome-euro.ucsc.edu/s/qferre/atyPeak_hg38>. We can see on this figure an example of rich CRE with many peaks, and a poorer CRE where many correlators for those TRs are missing which predictably has a lower score. As detailed previously, our approach estimates how “typical” each peak is, with respect to the usual combinations between sources (TRs and/or datasets) for a given cell line. As the model is unsupervised, anomaly score thresholds are at the user’s discretion. For example, a large scale analysis might exclude the lowest scoring peaks, but a focused study of a single or selected experimental series may specifically seek low-scoring peaks that might be caused by certain events of interest (mutations, etc.). It is also possible to use high-scoring peaks to detect CREs of interest and use that selection as a filter when looking at other genomic data, like we show here with ReMap 2020