The Authors present a comprehensive resource for the analysis of MeDIPseq data. Moreover, they apply it to carefully designed experiments and they release a dataset that is relevant for researchers interested in DNA methylation and cell differentiation.

Major Compulsory Revisions:

1. The MeDUSA computational pipeline is a big plus of this manuscript, and in particular its ability to identify differentially methylated regions (DMRs) and to point to nonCG methylated regions. The Authors indicate it is available upon request. In my experience this strongly inhibits people willing to use it. I would definitely suggest its public release, together with some documentation about installation and usage.

2. Regarding the identified DMRs, it would be very useful and more convincing having them represented in a heatmap over the compared samples. In addition I would also ask for a comparison between the DMRs that could be identified with the MeDIPs software.

3. Regarding the identification of nonCG methylation, I am somehow surprised about the identification of regions showing the presence of this mark and the absence of CG methylation. The combined presence of mC in both CG and nonCG contexts, together with the lower prevalence and lower methylation level characteristic of nonCG mC, might easily obscure the nonCG methylation pattern. Could the Authors comment on this and compare the predicted nonCG methylated regions with base resolution RRBS or BSseq data?

Minor Essential Revisions

1. do the Authors foresee any problem on the adopted normalization strategy due to the presence of nonCG methylation? In the case there is the same amount of mC sites in the CG context in two samples, and only one sample has also a lot of mC in the nonCG context, shouldn't this affect the normalization and cause an underestimation of the methylation level in the CG context?

2. the high number of identified DMRs is quite surprising. It would be interesting to have some statistics about their size and relative distance.

3. some more explanation would be nice about fragment length normalization.

4. the relevance and implication of the association between hyper-methylated regions with both up and down modulation of gene expression are quite obscure.

Level of interest: An article of importance in its field

Quality of written English: Acceptable

Statistical review: No, the manuscript does not need to be seen by a statistician.

Declaration of competing interests: I declare that I have no competing interests

Dear Editor,

the authors commented on all raised points by fulfilling most of the criticisms or adding clarifications in the text. For these reasons I suggest the publication of the manuscript on GigaScience

Level of interest: An article of importance in its field

Quality of written English: Acceptable

Statistical review: No, the manuscript does not need to be seen by a statistician.