Dear Dr. Nebert, this is a very interesting paper – thank you.
I agree with the authors, for most part. In addition to the weakness that they reported, I think DIP-seq has its inherent drawbacks. Without bisulfite modification, the methylation status is indirectly detected via antibody binding – and averaged throughout fragmented DNA. As a result, the detection resolution and sensitivity are low, or poor, making the data less informative. My hands-on experience also found that the downstream-validation work is difficult and time-consuming.
This is why, when I accepted the Core’s position 5 years ago, immediately I discontinued the DIP-seq service and discarded all those kits and antibodies. Despite the pressure I received from multiple researchers, I rejected their requests to perform DIP-seq for them. If you visit our website (http://med.uc.edu/eh/cores/genomics/services; – soon we will migrate to a new website, – thanks to Dr. Greis), you will see that this type of sequencing is not on our list.
I disagree with the authors [see original report below] that we should continue to use DIP-seq, because I wonder how accurate any “subtraction method” can be. Different antibodies bind DNA differently, and different labs use different experimental conditions. Whereas the authors highlighted the importance of experimental validation, unfortunately their subtraction method is not widely validated either.
The rapid progress in Next Generation Sequencing (NGS; also called whole-genome sequencing, WGS) methodology has made NGSt so fast and efficient – that we now have much better methods (whole-genome bisulfite-seuencing, WGBS; targeted methyl-seq; methylationEPIC Bead Chip; reduced representation bisulfite-seuencing, RRBS; NGS-based multiplex bisulfite PCR-sequencing; etc.) to study DNA methylation at much lower cost as well as with much higher quality/resolution. Today I am of the opinion that any method of DNA-methylation studies should follow up on the original gold standard (bisulfite modification) by also presenting data at single-CpG-site resolution.
By the way, one thing I like about NGS is that: the person who prepares the DNA samples is unlikely to conduct the sequencing, and the person who conducts the sequencing is unlikely to analyze the data. In addition, these days the sequencing data are required to be made publicly available – if one wishes to get the work published in a reputable journal. These factors make data-fabrication exceedingly difficult! 🙂