The final countdown (for the Hawaii paper)

Two years and ten days later…I’m ready to finish off this paper. In the meantime, I’ve discussed the paper with Steven a few times and decided on the following:

• We gotta go back to methylKit results. While the DSS methods seemed useful and are what I used for my dissertation, just spot-checking what the algorithm identified as DML did not make sense. I also can’t set a methylation threshold for DML with DSS, which makes it trickier to interpret or compare with other studies.
• I had Sam run EpiDiverse/snp with the Hawaii data, thinking that I could compare C->T SNP identification between methods. However, the EpiDiverse/snp output doesn’t provide a list of C->T SNPs. Sam and Steven are trying to troubleshoot this with the CEABIGR data, but for now I’m going to stick to BS-Snper output for SNP identification. The EpiDiverse/snp output information, however, could give us genotypic information that we could incorporate later on, since Maria was unable to tell us if the diploid female and tetraploid male oysters used were from related lines, or how many half- or full-sibling families were used for the triploid oysters I eventually got.
• A few methodological novelties I want to try with this dataset include a randomization test with methylKit, integration with C. gigas ATAC-Seq, csRNA-Seq, and 5’-GRO data, and KOG-MWU for DML comparison with other Crassostrea spp. epigenetic studies.
• Figure out if we can get the pH and water quality data

Original methylKit results

First things first, I wanted to find my original methylKit results. Thankfully, I didn’t delete anything from the Github repository (and if I did, I could always go back to a different version). How people ever find anything without Github, an online lab notebook, and large file storage with web links is something I will never understand.

To remind myself of what I did previously, I went through the paper and my lab notebook. I used a 25% cutoff to identify DML, which is different than the 50% I normally use. I also used min.per.group = 8L, which means a loci needs to have suitable coverage in eight samples per treatment.

The methylKit output is here, and I saved my .RData here. Turns out I never made figures with the methylKit version of the results, but I did find this lab notebook entry with the number of methylKit DML in each genome feature. My Jupyter notebook examining DML genomic location still has code that uses methylKit output. The numbers were consistent between the lab notebook and Jupyter notebook. However, I noticed that C->T SNPs were not removed prior to getting the count information! I removed the SNPs, then got updated counts. I needed to just modify a few lines of sed code at the end of the Jupyter notebook to get a table with the number of methylKit DML in each genome feature for contingency tests.

Table 1. Number of DML in each genome feature.

Genome Feature	pH DML (%)	Ploidy DML (%)	Common DML (%)
Total DML	34	24	1
Hypermethylated DML	24	8 (33.3%)	0
Hypomethylated DML	10	16 (66.6%)	1
Genes	28	20	1
Exon UTR	5	0	0
CDS	3	5	1
Introns	20	15	0
Upstream flanks	0	0	0
Downstream flanks	4	1	0
Intergenic regions	2	3	0
lncRNA	3	0	0
Transposable elements	15	6	0

• Create methylKit genome location counts
• Revise code to use methylKit output instead of DSS output
• Change color scheme

Randomization test

Going forward

1. Contingency tests for methylKit genome location
2. Revise code to use methylKit output instead of DSS output
3. Change color scheme for figures
4. methylKIt randomization test
5. Add Rajan’s comments to the Google Doc
6. Update methods
7. Update results
8. Revise discussion
9. Revise introduction
10. Transfer scripts used to a nextflow workflow