In PeteHaitch/methsim: An R package to simulate DNA methylation data

TODOs

Confirm validity of "2" classes and methods

Confirm that the new "2" classes (e.g., SimulateMethylome2, SimualteBSParam2) and methods (simulate2()) work as intended.

• [x] Empirical co-methylation model
• [x] $\beta$-values.
  • Looks great.
• [x] Correlation of $\beta$-values.
  • Way too correlated.
• [x] Within-fragment co-methylation.
  • Looks pretty good, probably upwardly biased due to inherent bias in LOR with small $d$ and extreme marginalProb.
• [x] Empirical co-methylation model with perturbed marginalProb.
• [x] $\beta$-values.
  • Looks great.
• [x] Correlation of $\beta$-values.
  • Reduced from above, but still probably too correlated.
• [x] Within-fragment co-methylation.
  • Looks pretty good, probably upwardly biased due to inherent bias in LOR with small $d$ and extreme marginalProb.
• [x] Independence model, i.e., $LOR = 0$ uniformly.
• [x] $\beta$-values.
  • Looks great.
• [x] Correlation of $\beta$-values.
  • Way too correlated.
• [x] Within-fragment co-methylation.
  • Looks pretty good, approximately flat as a function of $IPD$. However, there is clear evidence of bias because $LOR \approx 1$ meaning $OR \approx 2^1 = 2$.
• [x] Independence model, i.e., $LOR = 0$ uniformly, with perturbed marginalProb
• [x] $\beta$-values.
  • Looks great.
• [x] Correlation of $\beta$-values.
  • Reduced from above, but still probably too correlated.
• [x] Within-fragment co-methylation.
  • Looks pretty good, probably upwardly biased due to inherent bias in LOR with small $d$ and extreme marginalProb.

Once confirmed, deprecate the old "1" classes and methods in favour of the "2" classes and methods.

UP TO HERE

• [ ] Classes and methods deprecated.

PeteHaitch/methsim documentation built on May 8, 2019, 1:32 a.m.