Experimental and Computational Approaches for Non-CpG Methylation Analysis

Round 1

Reviewer 1 Report

The review addresses an area of research, which is overall neglected and indeed suffers from technical limitations and difficulties in analysis. The review is thus timely and relevant, however, there are important omissions that the authors need to address.

Main comments

The overall coverage of the historical techniques for the characterisation of non-CpG methylation is satisfactory, however, the bisulfite sequencing section lacks some major tools and practices in the bioinformatics analyses of non-CG methylation. Sections 2.3 and 3 need to be corrected and extended to add all relevant information.

The bioinformatics tool ‘bismark’ detects CG, CHG and CHH since its development in 2010 (https://www.bioinformatics.babraham.ac.uk/projects/bismark/) – output from Bismark is standardly used with the program SeqMonk (https://www.bioinformatics.babraham.ac.uk/projects/seqmonk/) to analyse, quantitate and visualise methylation in any context. These tools have been used with some of the very first WGBS datasets, which also focused on non-CG methylation. Moreover, bismark has tools designed specifically for the analysis of non-CG methylation, such as conversion artefact filtering. Non-conversion filtering was used with the first next generation coupled BS-sequencing protocols by A Meissner’s team (2008-2010), including for plants and high non-CG containing cells such as ES cells. Detailed analysis of artefacts and biases affecting non-CG methylation and its interpretation and quantitation in WGBS data has been performed in a landmark study, not mentioned here: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-018-1408-2

Also, same genome synthetic unmethylated controls rather than only Lambda DNA (https://www.pnas.org/doi/full/10.1073/pnas.1515937112) or validation with non-BS dependent methods (https://academic.oup.com/nar/article/48/22/12675/6017363?login=false) are good examples of evolved approaches applied to validate non-CG methylation results.

Overall, the bisulfite method has intrinsic noise in the non-CG context, due to incomplete conversion, which historically has led to overestimation of methylation in this context. Because of this revelation, the enzyme-based conversion method (EM-seq) has been consequently developed, which is specifically suited for non-CG analysis, and confirms non-CG levels are generally lower than reported with whole genome BS sequencing (https://epigeneticsandchromatin.biomedcentral.com/articles/10.1186/s13072-020-00361-9).

All of these are important developments specifically originating from and aiming to improve the analysis of non-CG methylation and are entirely omitted in this review.

 

Specific comments:

 

Lines 11-12, line 29 – non-CG methylation varies between cell types, it is either abundant (neurons, oocytes), moderate to low (ES cells) or nearly non-existent in most cell types; it will be more accurate if authors give a range rather than one figure as it raises questions what cell type this figure is for and if it is accurate at all.  

 

Line 127 – reference 28 is about hairpin-BS and not about lambda unmethylated control, authors to check and correct their references. Note, that the Genome Biology reference above addresses both Lambda and conversion controls as well as compares conventional cloning-based BS sequencing, hairpin and WGBS in relation to non-CG methylation output.

 

Section 2.3:

Line 139-140 – this is inaccurate explanation of what how methylation is calculated (it’s converted vs non-converted cytosines)

Lines 142-143 - hairpin BS sequencing HAS single base resolution

Lines 144-145 – WGBS has many drawbacks specifically for the detection of non-CG methylation, check the above references. It is a great technique (the best available before EM-seq was developed) but needs special controls and analysis regarding non-CG context due to high noise in non-CG context

Lines 151-155 – RRBS is heavily aimed at CG methylation and is not a good technique for non-CG methylation, as non-CG is low around CGIs, for which MspI-based RRBS enriches. Ziller et al did a good study because these datasets were already available and are cheaper (WGBS was too expensive in 2011), despite being a compromise regarding non-CG methylation. MspI-RRBS should not be presented as a technique suitable specifically for non-CG methylation analysis, the potential pluses and clear minuses should be explained objectively.

 

Line 158 – antibody-based enrichment – very vague, key in peak-based methods is that only asymmetric peaks are selected for (potentially) representing non-CG methylation, and this is also negatively affected by antibody’s preferences to bind to cytosine-rich asymmetric sequences – these are drawbacks, which are not mentioned at all.

 

Section 3 – extend with the omitted tools and non-conversion approaches as discussed in main comments. In the figure against the ‘aligner’ – is ‘BS-mark’ actually meant for ‘bismark’? There is no other aligner called ‘BS-mark’. Correct and add relevant citations such as: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3102221/

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

"Experimental and computational approaches for non-CpG methylation analysis" by Ramasamy et al. is a brief review describing the current methods available for detecting and analyzing non-CpG methylation. The authors briefly discuss the limited knowledge available about non-CpG methylation and then examine both the wetlab techniques used to ascertain non-CpG methylation data and the bioinformatic tools currently developed to analyze these data. The authors go on to describe the likely next steps needed, which are primarily to develop bioinformatic tools to filter WGBS data for non-CpG methylation.

 

Non-CpG methylation is a very understudied area of epigenomics, so publishing a review like this is worthwhile. The manuscript has some awkward English that needs to be rectified, as there are some sentences that I struggle to understand. The review also feels a bit cursory to me, but this is probably because the tools available for non-CpG methylation characterization are limited, to which the authors allude. I have only minor comments:

 

1. Consider cleaner header fonts in Figure 1. The current font type appears blurry and low resolution.

2. Line 28-29 - Maybe use "conversely" rather than "instead."

3. I don't understand this sentence in line 51-52: "The methods designed to specifically identify..."

4. Line 63, I would replace "besides" with "additionally."

 

Author Response

Please see the attachment

Author Response File: Author Response.docx