ChIP-seq is used by thousands of research studies to profile histone modifications in cancer. However, methods developed for normal diploid genomes, when applied to cancer samples, can result in false discoveries due to the presence of copy number aberrations distorting the ChIP-seq signal. In order to circumvent this issue, our group has developed a set of ChIP-seq data analysis methods for cancer (boevalab.com/tools.html), including • HMCan: a method to call ChIP-seq peaks and normalize read density profiles for copy number bias [1], • LILY: a method to identify super-enhancer regions in cancer [2].
We applied HMCan and LILY to detect super-enhancer regions in 25 neuroblastoma cell lines and 6 patient derived mouse xenografts. Analysis of super-enhancer landscape in these samples suggested that neuroblastoma cells can be in two different epigenetic and transcriptional states. Single cell analysis showed that both states can co-exist in the same patient. One state was associated with amplification or high expression of the MYCN oncogene and high activity of noradrenergic transcription factors: PHOX2B, GATA3 and HAND2. The second state was characterized by the high activity of the AP-1 transcription complex and transcription factors like PRRX1 and RUNX1. Furthermore, we demonstrated that cells in the second, neural crest-like state were more resistant to chemotherapy [2].
References:
1. Ashoor et al. Bioinformatics, 2013, 29(23): 2979-2986
2. Boeva et al. Nature Genetics, 2017, 49(9):1408-1413