Evaluation of ChIC-based data requires normalization that properly retains signal-to-noise ratios

Abstract

Several chromatin immunocleavage-based (ChIC) methods using Tn5 transposase have been developed to profile histone modifications and transcription factors bindings1-5. A recent preprint by Wang et al. raised potential concerns that these methods are prone to open chromatin bias6. While the authors are appreciated for alerting the community for this issue, it has been previously described and discussed by Henikoff and colleagues in the original CUT&Tag paper3,7. However, as described for CUT&Tag3, the signal-to-noise ratio is essential for Tn5-based profiling methods and all antibody-based enrichment assays. Based on this notion, we would like to point out a major analysis issue in Wang et al. that caused a complete loss or dramatic reduction of enrichment at true targets for datasets generated by Tn5-based methods, which in turn artificially enhanced the relative enrichment of potential open chromatin bias. Such analysis issue is caused by distinct background normalizations used towards ChIP-based (chromatin immunoprecipitation) data and Tn5-based data in Wang et al. Only the normalization for Tn5-based data, but not ChIP-seq based data, yielded such effects. Distortion of such signal-to-noise ratio would consequently lead to misleading results.