I am dealing for the first time with some samples which have undergone the Cut&Run procedure rather than the traditional ChIP-seq.
In the past couple of years, I have only processed standard ChIP-seq samples so I wondered if, regarding the data analysis pipeline, there are differences I should be aware of before processing the data (specific tools or parameters for trimming, alignment, etc.).
Basically it is the same. The only difference is that C+R can produce some very small fragments so if this is paired-end sequencing then some people tune aligners such as bowtie2 a bit, see e.g. Cut & run - bowtie2 dovetail option but I am not sure whether this is truely necessary. For peak calling there a some C+R-specific callers such as SEACR that claim to perform better with very high quality data which have extremely low levels of noise, hence might overinterpret small peaks. Anecdotically I heard that many use macs2 though as often C+R quality is just the same as ChIP-seq, and the few C+R datasets I had a look at worked quite ok with macs2. Once you have your peaks it is the same as in any other (differential) analysis if you ask me. That all is just my "thinking aloud" so take it with a grain of salt. I mean after all it is pulldown of protein-bound DNA and then sequencing, there is no fundamental difference between the two assays other than that ChIP does crosslinking and sonication on extracted chromatin while C+R uses MNase-based cutting without crosslinking on permeabilzed cells. It is still an antibody-based assay so quality widely varies and depends on epitope uniqueness, abundance and the daily mood of the antibody god. In my limited experience C+R, while a lot more convenient on the lab side as it is super fast and required few cells, is not a guarantee for a good quality dataset (not that you were asking about that but I felt like writing about it :) )
Thank you so much for your extensive and frank comment in this regard, I appreciate it!
I don't have paired-end data, so I guess I do not need to deal with some bowtie2 tuning..I ran bowtie2 the same way I was used to running it for ChIP-seq and it worked great.
For the peak calling, I will try it with SEACR and compare the results with MACS2. As long as MACS2 works well, I always prefer to stick to tools I am familiar with and that I trust due to past experience.
In my hands, SEACR overcalls wildly unless your dataset is pristine quality and literally almost zero background. And it really has almost no parameters you can tweak, so we've found it easier to play with MACS2 to get a reasonable peakset.
Thank you so much for your extensive and frank comment in this regard, I appreciate it! I don't have paired-end data, so I guess I do not need to deal with some
bowtie2tuning..I ranbowtie2the same way I was used to running it for ChIP-seq and it worked great.For the peak calling, I will try it with
SEACRand compare the results withMACS2. As long asMACS2works well, I always prefer to stick to tools I am familiar with and that I trust due to past experience.In my hands, SEACR overcalls wildly unless your dataset is pristine quality and literally almost zero background. And it really has almost no parameters you can tweak, so we've found it easier to play with MACS2 to get a reasonable peakset.
Thanks a lot for your suggestion!