Revise DNA Methylation tutorial

[sr320]

Sam Bogan Review of “DNA Methylation Analyses” by Steven Roberts, Sam White, and Yaamini Venkataraman link: https://d.pr/f/332Q5v Original text is in blue. Suggested text is in red.

General Feedback

• [x] I love linked resources regarding reviews, BS-seq method comparisons, RRBS specifications etc.
• [x] Signposts in the tutorial emphasizing forks in the road for WGBS vs RRBS are extremely useful.
• [x] Some areas could use more explanation for first time users, such as explaining terms associated with Bismark and Bowtie. I’ve provided some examples below.
• [x] It would help to highlight some parameters that, when changed, can yield significant effects. The code chunks associated with genome prep and alignment would be great places for these signposts. I’ve highlighted a couple parameters below.
• [ ] The page would benefit from one or two diagnostic plots to encourage readers to examine their data. Such plots could be (i) PCAs of BS-seq libraries or (ii) plots of read coverage across CpGs and/or gene bodies.

Introduction

• [x] It would be helpful to briefly define DNA methylation in the first paragraph.
• [x] The intro says the tutorial focuses on loci- and gene-level methylation, but methylation is not quantified across gene bodies in the below code as far as I can tell. It would be worth adding an example of how to do this!
• [x] The preface to conversion of unmethylated bases could include enzymatic conversion (EM-seq) since folks are starting to use this method and it can integrate with BS-seq pipelines. “…methylation in bisulfite-treated DNA sequence data” could be edited to “methylation in bisulfite-treated and enzymatically converted DNA sequence data.”
• [x] The last sentence of the intro could add “WGBS reads were derived from a species of marine invertebrate” with any explanation of why the species was chosen, if appropriate.
• [x] If you’re open to readers locating and using the data, it’s worth listing the species and SRA.
• [x] It would benefit readers to list what workflow steps are covered in the tutorial and that differential methylation is not discussed.

Sequence Quality

• [x] “…most marine invertebrates demonstrating…” should be changed to “…most invertebrates demonstrating…”.
• [x] “… see something similar to this” should be changed to “… see fastQC per base sequence content similar the plot shown below.”

Read Alignment

• [x] Bismark should be cited.
• [x] “Thus content below will provided with the assumption…” looks like a type and could be changed to “The content below assumes…”
• [x] It would be helpful to explain the tasks genome preparation performs (e.g. in silico bisulfite conversion of the reference genome, pre-filtering certain regions prior to alignment, etc).
• [x] Highlighting which of these steps is most significant if they are or aren’t performed. For example, pre-filtering/masking with this step can be important. Generally, masking C-T SNPs would be a useful discussion point to add.
• [x] “You should expect to prepared genome with new directories including” looks like a typo and could also be more specific to read “Bismark will create the following subdirectories as outputs of the above code for genome preparation.”
• [x] All instances of “BAM”/”bam” should be upper case, or in a consistent format.
• [x] The example code structures for alignment should contain comments/annotations for the defined parameters. This doesn’t seem as necessary for the deduplication code.
• [ ] Defining thresholds for/the importance of alignment minimum score minimums may be useful considering reduced alignment efficiency in BS-seq data. Whatever parameter you think is most important to highlight.

Methylation Quantification

• [x] Add comments/annotations to lines of the first and fourth code chunks in this section.
• [x] It would be helpful to set expectations regarding how methylation proportions vary (e.g., expect bimodal distributions for certain features and what kinds of distributions to investigate further).

[sr320]

Review from Jason Johns: https://d.pr/f/zABmGv

[yaaminiv]

More things to add

• [ ] sequencing depth considerations (@sr320 assigned)
• [x] example mapping report from Bismark (@yaaminiv)
• [x] subsetting data/looking at alignment sensitivity
• [x] add vertebrate examples (kathryn)
• [ ] add code descriptions for file conversion section
• [ ] add code descriptions for post-SNP ID section
• [ ] resources for downstream applications
• [ ] update .bib file with references

[yaaminiv]

@sr320 this part of the SNP ID tutorial doesn't make sense to me:

---

If for example in downstream analysis you would might be using a _5x.tab file described above the following is a means in R to remove these SNPs.

# Read in CT SNP file
ct <- read.csv("../output/CT-SNP.vcf", header = FALSE, sep = "\t") %>%
  mutate(loci = paste0(V1, "_", V2))

# 1. List all files with _5x.tab suffix
files <- list.files(path = "../data/", pattern = "_5x.tab$", full.names = TRUE)

# 2. Iterate over each file
for(file in files) {

  # Extract base filename without the directory for naming purposes
  base_name <- basename(file)

  # Read the file
  data <- read.csv(file, header = FALSE, sep = "\t")

  # Modify the data
  modified_data <- data %>%
    mutate(loci = paste0(V1, "_", V2)) %>%
    anti_join(ct, by = "loci") %>%
    select(-loci)

  # Write the modified data to an output file
  output_file <- paste0("../output/f", base_name)
  write.table(modified_data, file = output_file, sep = "\t", row.names = FALSE, quote = FALSE, col.names = FALSE)
}

[sr320]

completed and published