DML Analysis Part 32

Based on feedback from Katie, Steven, and Alan, I went back to my code and revised C. virginica gonad methylation trends.

Revised timeline

But first…let’s revist my grand procalamtion: “I’m going to finish this paper by the end of the month.” (lol) I think I can finish my analyses by the end of this week and have some sort of draft discussion. Before the next E2O meeting, I’ll definitely have a draft paper ready.

Characterizing loci locations

In this Jupyter notebook, I characterized the location of all 5x CpGs I had data for, sparsely methylated loci, and unmethyalted loci. Katie suggested I make a table with this information for statistical analyses, so I did.

Table 1. Locations of 5x CpGs enriched by MBD treatment, methylated loci, sparsely methylated loci, and unmethylated loci. “Other” refers to any loci that did not overlap with exons, introns, transposable elements (all), and putative promoters.

Category All 5x CpGs Methylated Sparsely methylated Unmethylated
Total 4,304,257 3,181,904 481,788 640,565
Unique genes 54,619 44,505 47,243 47,584
mRNA coding regions 3,140,744 2,437,901 303,890 398,953
Exons 1,366,779 1,013,691 105,871 247,217
Introns 1,811,271 1,448,786 201,553 160,932
Transposable elements (all) 1,011,883 755,222 155,293 101,368
Transposable elements (Cg) 767,604 610,208 108,858 48,538
Putative promoters 203,376 134,534 27,443 41,399
Other 627,257 386,003 86,923 154,331

Chi-squared tests

For this round of chi-squared tests, I updated my overlap proportions file. I used this code to conduct chi-squared tests for various groupings. Every single time, I found that my distributions were significantly different. However, I’m now doubting how effective chi-squared tests are as a statistical approach. Looking at the observed, expected, and residual values from chisq.test from my test comparing all CG motifs with those enriched by MBD, I notice that my expected values are not what I want. In comparing these distributions, I want the background proportions (in this case, all CG motifs), to serve as the expected values, but that’s not the case:

Screen Shot 2019-04-29 at 7 14 20 PM

I psoted this issue to get some clarification on my methods.

Revised figures

In the meantime, I revised my figures!


Figure 1. Distribution of CpGs enriched by MBD versus all CpGs in the C. virginica genome.


Figure 2. Distribution of methylated CpGs versus those enriched by MBD.


Figure 3. Distribution of methylated, sparsely methylated, and unmethylated CpGs versus CpGs enriched by MBD.


Figure 4. Distribution of DML versus methylated CpGs.

Once I sort out my statstical tests I can actually add significances to my figures.

Going forward

  1. Sort out statistical tests
  2. Describe (somehow) genes with DML in them
  3. Figure out what’s going on with the gene background
  4. Figure out what’s going on with DMR
  5. Work through gene-level analysis
  6. Update paper repository
  7. Start writing the discussion
Written on April 29, 2019